Medical Physics

Pseudo‐CT synthesis in adaptive radiotherapy based on a stacked coarse‐to‐fine model: Combing diffusion process and spatial‐frequency convolutions

AbstractBackgroundCone beam computed tomography (CBCT) provides critical anatomical information for adaptive radiotherapy (ART), especially for tumors in the pelvic region that undergo significant deformation. However, CBCT suffers from inaccurate Hounsfield Unit (HU) values and lower soft tissue contrast. These issues affect the accuracy of pelvic treatment plans and implementation of the treatment, hence requiring correction.PurposeA novel stacked coarse‐to‐fine model combining Denoising Diffusion Probabilistic Model (DDPM) and spatial‐frequency domain convolution modules is proposed to enhance the imaging quality of CBCT images.MethodsThe enhancement of low‐quality CBCT images is divided into two stages. In the coarse stage, the improved DDPM with U‐ConvNeXt architecture is used to complete the denoising task of CBCT images. In the fine stage, the deep convolutional network model jointly constructed by fast Fourier and dilated convolution modules is used to further enhance the image quality in local details and global imaging. Finally, the accurate pseudo‐CT (pCT) images consistent with the size of the original data are obtained. Two hundred fifty paired CBCT‐CT images from cervical and rectal cancer, combined with 200 public dataset cases, were used collectively for training, validation, and testing.ResultsTo evaluate the anatomical consistency between pCT and real CT, we have used the mean(std) of structure similarity index measure (SSIM), peak signal to noise ratio (PSNR), and normalized cross‐correlation (NCC). The numerical results for the above three metrics comparing the pCT synthesized by the proposed model against real CT for cervical cancer cases were 87.14% (2.91%), 34.02 dB (1.35 dB), and 88.01% (1.82%), respectively. For rectal cancer cases, the corresponding results were 86.06% (2.70%), 33.50 dB (1.41 dB), and 87.44% (1.95%). The paired t‐test analysis between the proposed model and the comparative models (ResUnet, CycleGAN, DDPM, and DDIM) for these metrics revealed statistically significant differences (p < 0.05). The visual results also showed that the anatomical structures between the real CT and the pCT synthesized by the proposed model were closer. For the dosimetric verification, mean absolute error of dosimetry (MAEdoes) values for the maximum dose (Dmax), the minimum dose (Dmin), and the mean dose (Dmean) in the planning target volume (PTV) were analyzed, with results presented as mean (lower quartile, upper quartile). The experimental results show that the values of the above three dosimetry indexes (Dmin, Dmax, and Dmean) for the pCT images synthesized by the proposed model were 0.90% (0.48%, 1.29%), 0.82% (0.47%, 1.17%), and 0.57% (0.44%, 0.67%). Compared with 10 cases of the original CBCT image by Mann–Whitney test (p < 0.05), it also proved that pCT can significantly improve the accuracy of HU values for the dose calculation.ConclusionThe pCT synthesized by the proposed model outperforms the comparative models in numerical accuracy and visualization, promising for ART of pelvic cancers.

A novel network architecture for post‐applicator placement CT auto‐contouring in cervical cancer HDR brachytherapy

Abstract Background High‐dose‐rate brachytherapy (HDR‐BT) is an integral part of treatment for locally advanced cervical cancer, requiring accurate segmentation of the high‐risk clinical target volume (HR‐CTV) and organs at risk (OARs) on post‐applicator CT (pCT) for precise and safe dose delivery. Manual contouring, however, is time‐consuming and highly variable, with challenges heightened in cervical HDR‐BT due to complex anatomy and low tissue contrast. An effective auto‐contouring solution could significantly enhance efficiency, consistency, and accuracy in cervical HDR‐BT planning. Purpose To develop a machine learning‐based approach that improves the accuracy and efficiency of HR‐CTV and OAR segmentation on pCT images for cervical HDR‐BT. Methods The proposed method employs two sequential deep learning models to segment target and OARs from planning CT data. The intuitive model, a U‐Net, initially segments simpler structures such as the bladder and HR‐CTV, utilizing shallow features and iodine contrast agents. Building on this, the sophisticated model targets complex structures like the sigmoid, rectum, and bowel, addressing challenges from low contrast, anatomical proximity, and imaging artifacts. This model incorporates spatial information from the intuitive model and uses total variation regularization to improve segmentation smoothness by applying a penalty to changes in gradient. This dual‐model approach improves accuracy and consistency in segmenting high‐risk clinical target volumes and organs at risk in cervical HDR‐BT. To validate the proposed method, 32 cervical cancer patients treated with tandem and ovoid (T&O) HDR brachytherapy (3–5 fractions, 115 CT images) were retrospectively selected. The method's performance was assessed using four‐fold cross‐validation, comparing segmentation results to manual contours across five metrics: Dice similarity coefficient (DSC), 95% Hausdorff distance (HD 95 ), mean surface distance (MSD), center‐of‐mass distance (CMD), and volume difference (VD). Dosimetric evaluations included D90 for HR‐CTV and D2cc for OARs. Results The proposed method demonstrates high segmentation accuracy for HR‐CTV, bladder, and rectum, achieving DSC values of 0.79 ± 0.06, 0.83 ± 0.10, and 0.76 ± 0.15, MSD values of 1.92 ± 0.77 mm, 2.24 ± 1.20 mm, and 4.18 ± 3.74 mm, and absolute VD values of 5.34 ± 4.85 cc, 17.16 ± 17.38 cc, and 18.54 ± 16.83 cc, respectively. Despite challenges in bowel and sigmoid segmentation due to poor soft tissue contrast in CT and variability in manual contouring (ground truth volumes of 128.48 ± 95.9 cc and 51.87 ± 40.67 cc), the method significantly outperforms two state‐of‐the‐art methods on DSC, MSD, and CMD metrics ( p ‐value < 0.05). For HR‐CTV, the mean absolute D90 difference was 0.42 ± 1.17 Gy ( p ‐value > 0.05), less than 5% of the prescription dose. Over 75% of cases showed changes within ± 0.5 Gy, and fewer than 10% exceeded ± 1 Gy. The mean and variation in structure volume and D2cc parameters between manual and segmented contours for OARs showed no significant differences ( p ‐value > 0.05), with mean absolute D2cc differences within 0.5 Gy, except for the bladder, which exhibited higher variability (0.97 Gy). Conclusion Our innovative auto‐contouring method showed promising results in segmenting HR‐CTV and OARs from pCT, potentially enhancing the efficiency of HDR BT cervical treatment planning. Further validation and clinical implementation are required to fully realize its clinical benefits.

Attention 3D U‐NET for dose distribution prediction of high‐dose‐rate brachytherapy of cervical cancer: Direction modulated brachytherapy tandem applicator

AbstractBackgroundDirection Modulated Brachytherapy (DMBT) enables conformal dose distributions. However, clinicians may face challenges in creating viable treatment plans within a fast‐paced clinical setting, especially for a novel technology like DMBT, where cumulative clinical experience is limited. Deep learning‐based dose prediction methods have emerged as effective tools for enhancing efficiency.PurposeTo develop a voxel‐wise dose prediction model using an attention‐gating mechanism and a 3D UNET for cervical cancer high‐dose‐rate (HDR) brachytherapy treatment planning with DMBT six‐groove tandems with ovoids or ring applicators.MethodsA multi‐institutional cohort of 122 retrospective clinical HDR brachytherapy plans treated to a prescription dose in the range of 4.8–7.0 Gy/fraction was used. A DMBT tandem model was constructed and incorporated onto a research version of BrachyVision Treatment Planning System (BV‐TPS) as a 3D solid model applicator and retrospectively re‐planned all cases by seasoned experts. Those plans were randomly divided into 64:16:20 as training, validating, and testing cohorts, respectively. Data augmentation was applied to the training and validation sets to increase the size by a factor of 4. An attention‐gated 3D UNET architecture model was developed to predict full 3D dose distributions based on high‐risk clinical target volume (CTVHR) and organs at risk (OARs) contour information. The model was trained using the mean absolute error loss function, Adam optimization algorithm, a learning rate of 0.001, 250 epochs, and a batch size of eight. In addition, a baseline UNET model was trained similarly for comparison. The model performance was evaluated on the testing dataset by analyzing the outcomes in terms of mean dose values and derived dose‐volume‐histogram indices from 3D dose distributions and comparing the generated dose distributions against the ground‐truth dose distributions using dose statistics and clinically meaningful dosimetric indices.ResultsThe proposed attention‐gated 3D UNET model showed competitive accuracy in predicting 3D dose distributions that closely resemble the ground‐truth dose distributions. The average values of the mean absolute errors were 1.82 ± 29.09 Gy (vs. 6.41 ± 20.16 Gy for a baseline UNET) in CTVHR, 0.89 ± 1.25 Gy (vs. 0.94 ± 3.96 Gy for a baseline UNET) in the bladder, 0.33 ± 0.67 Gy (vs. 0.53 ± 1.66 Gy for a baseline UNET) in the rectum, and 0.55 ± 1.57 Gy (vs. 0.76 ± 2.89 Gy for a baseline UNET) in the sigmoid. The results showed that the mean absolute error (MAE) for the bladder, rectum, and sigmoid were 0.22 ± 1.22 Gy (3.62%) (p = 0.015), 0.21 ± 1.06 Gy (2.20%) (p = 0.172), and ‐0.03 ± 0.54 Gy (1.13%) (p = 0.774), respectively. The MAE for D90, V100%, and V150% of the CTVHR were 0.46 ± 2.44 Gy (8.14%) (p = 0.018), 0.57 ± 11.25% (5.23%) (p = 0.283), and ‐0.43 ± 19.36% (4.62%) (p = 0.190), respectively. The proposed model needs less than 5 s to predict a full 3D dose distribution of 64 × 64 × 64 voxels for any new patient plan, thus making it sufficient for near real‐time applications and aiding with decision‐making in the clinic.ConclusionsAttention gated 3D‐UNET model demonstrated a capability in predicting voxel‐wise dose prediction, in comparison to 3D UNET, for DMBT intracavitary brachytherapy planning. The proposed model could be used to obtain dose distributions for near real‐time decision‐making before DMBT planning and quality assurance. This will guide future automated planning, making the workflow more efficient and clinically viable.

Simultaneous catheter and multicriteria optimization for HDR cervical cancer brachytherapy with a complex intracavity/interstitial applicator

AbstractBackgroundComplex intracavity and interstitial (IC/IS) applicators, such as the Venezia applicator, can improve the HR‐CTV coverage while adequately protecting organs at risk in the treatment of cervical cancer with high‐dose‐rate (HDR) brachytherapy. Although the Venezia applicator offers more choice for catheter selection, commercially available catheter and dose optimization algorithms are still missing for complex applicators. Moreover, studies on catheter and dose optimization for IC/IS implants in the treatment of cervical cancer are still limited.PurposeThis work aims to combine a GPU‐based multi‐criteria optimization (gMCO) algorithm with a sparse catheter (SC) optimization algorithm for the Venezia applicator.MethodsFifty‐eight cervical cancer patients who received 28 Gy in 4 fx of HDR brachytherapy with the Venezia applicator (combination to external beam radiation therapy) are retrospectively revisited. The modelization of the applicator is done by virtually reconstructing all the IS catheters passing through the ring. Template catheters are reconstructed using an in‐house python script. To perform simultaneous MCO and SC optimization (SC+MCO), the objective function includes aggregated dose objectives in a weighted sum and a group sparsity term that individually penalizes the contribution of IS catheters. Plans generated with the SC+MCO algorithm are compared with plans generated with MCO using clinical catheters (CC+MCO) and the clinical plans (CP). The EMBRACE II soft constraints (planning aims) and hard constraints (limits for prescribed dose) are used as plan evaluation criteria.ResultsCC+MCO gives the most important gain with an increase up to 20.7% in meeting all EMBRACE II soft constraints compared with CP. The SC+MCO algorithm (adding catheter optimization to MCO) provides a second order increase (up to 12.1% with total acceptance rate of 60.3% or 35/58) in the acceptance rate versus CC+MCO (total increase of 32.8% vs. CP). Acceptance rate in EMBRACE II hard constraints is 98.3% (57/58) for both CC+MCO and SC+MCO versus 91.4% (53/58) for CP. The median SC+MCO optimization time is 11 s to generate a total of 5000 Pareto‐optimal plans with different catheter configurations (position and number) for each fraction.ConclusionsSimultaneous catheter and MCO optimization is clinically feasible for HDR cervical cancer brachytherapy using the Venezia applicator. Clinical catheter configurations could be improved and/or the catheter number could be reduced without decreasing plan quality using SC+MCO compared with the CP.

Slice‐prompted HR‐CTV interactive segmentation for cervical cancer brachytherapy: A multi‐center study

Abstract Background In computed tomography (CT)‐guided cervical cancer brachytherapy, the manual contouring for the high‐risk clinical target volume (HR‐CTV) is a time‐consuming and expertise‐dependent process. Furthermore, automated approaches struggle with ambiguous boundaries of HR‐CTV. Purpose We aimed to develop a clinically efficient interactive segmentation framework integrating deep learning with clinician expertise. Methods and materials We propose a slice‐prompted interactive segmentation method (SPSeg) for HR‐CTV delineation in CT‐guided cervical cancer brachytherapy. Clinicians provided sparse prompts by manually outlining HR‐CTV on key slices, which were then encoded into a 3D U‐Net architecture to guide full‐volume segmentation. We investigated two architectural variants: SPSeg‐Mono, which jointly processes the CT images and the prompt masks with a single encoder; and SPSeg‐Dual, which employs two separate encoders for image and prompt, fusing their features at a deeper level. The model was trained on 640 CT scans (from 160 patients) and validated on 160 scans (40 patients) from a single center, and externally tested on three multi‐center cohorts: 400 scans (100 patients), 115 scans (40 patients), and 150 scans (30 patients), respectively. Evaluation included Dice Similarity Coefficient (DSC), 95% Hausdorff Distance (HD95), a 5‐point Likert scale for clinical acceptability, time efficiency, and inter‐observer agreement. Results Performance consistently improved with the addition of prompt slices, with SPSeg‐Dual outperforming SPSeg‐Mono. Without prompts, the model yielded DSCs of 0.83, 0.76, and 0.76, and HD95s of 7.5, 10.1, and 11.6 mm for Test Sets 1, 2, and 3, respectively. With the addition of just three prompt slices, DSCs increased significantly to 0.95, 0.92, and 0.91, while HD95s decreased to 2.1, 3.1, and 3.2 mm, respectively (all p  < 0.001). Qualitative scores confirmed high clinical acceptability (mean Likert scores > 3), and the interactive method substantially reduced contouring time for both clinicians (from 11.7 to 1.7 min for Clinician A, and from 9.9 to 1.5 min for Clinician B). It also improved inter‐observer agreement, with DSC increasing from 0.88 to 0.93 and HD95 decreasing from 3.2 to 2.5 mm ( p  < 0.001). Conclusions The proposed SPSeg method effectively integrates clinical expertise with deep learning, offering a highly precise and efficient solution for HR‐CTV delineation in cervical cancer brachytherapy.

Neural network dose prediction for cervical brachytherapy: Overcoming data scarcity for applicator‐specific models

AbstractBackground3D neural network dose predictions are useful for automating brachytherapy (BT) treatment planning for cervical cancer. Cervical BT can be delivered with numerous applicators, which necessitates developing models that generalize to multiple applicator types. The variability and scarcity of data for any given applicator type poses challenges for deep learning.PurposeThe goal of this work was to compare three methods of neural network training—a single model trained on all applicator data, fine‐tuning the combined model to each applicator, and individual (IDV) applicator models—to determine the optimal method for dose prediction.MethodsModels were produced for four applicator types—tandem‐and‐ovoid (T&O), T&O with 1–7 needles (T&ON), tandem‐and‐ring (T&R) and T&R with 1–4 needles (T&RN). First, the combined model was trained on 859 treatment plans from 266 cervical cancer patients treated from 2010 onwards. The train/validation/test split was 70%/16%/14%, with approximately 49%/10%/19%/22% T&O/T&ON/T&R/T&RN in each dataset. Inputs included four channels for anatomical masks (high‐risk clinical target volume [HRCTV], bladder, rectum, and sigmoid), a mask indicating dwell position locations, and applicator channels for each applicator component. Applicator channels were created by mapping the 3D dose for a single dwell position to each dwell position and summing over each applicator component with uniform dwell time weighting. A 3D Cascade U‐Net, which consists of two U‐Nets in sequence, and mean squared error loss function were used. The combined model was then fine‐tuned to produce four applicator‐specific models by freezing the first U‐Net and encoding layers of the second and resuming training on applicator‐specific data. Finally, four IDV models were trained using only data from each applicator type. Performance of these three model types was compared using the following metrics for the test set: mean error (ME, representing model bias) and mean absolute error (MAE) over all dose voxels and ME of clinical metrics (HRCTV D90% and D2cc of bladder, rectum, and sigmoid), averaged over all patients. A positive ME indicates the clinical dose was higher than predicted. 3D global gamma analysis with the prescription dose as reference value was performed. Dice similarity coefficients (DSC) were computed for each isodose volume.ResultsFine‐tuned and combined models showed better performance than IDV applicator training. Fine‐tuning resulted in modest improvements in about half the metrics, compared to the combined model, while the remainder were mostly unchanged. Fine‐tuned MAE = 3.98%/2.69%/5.36%/3.80% for T&O/T&R/T&ON/T&RN, and ME over all voxels = –0.08%/–0.89%/–0.59%/1.42%. ME D2cc were bladder = –0.77%/1.00%/–0.66%/–1.53%, rectum = 1.11%/–0.22%/–0.29%/–3.37%, sigmoid = –0.47%/–0.06%/–2.37%/–1.40%, and ME D90 = 2.6%/–4.4%/4.8%/0.0%. Gamma pass rates (3%/3 mm) were 86%/91%/83%/89%. Mean DSCs were 0.92%/0.92%/0.88%/0.91% for isodoses ≤ 150% of prescription.Conclusions3D BT dose was accurately predicted for all applicator types, as indicated by the low MAE and MEs, high gamma scores and high DSCs. Training on all treatment data overcomes challenges with data scarcity in each applicator type, resulting in superior performance than can be achieved by training on IDV applicators alone. This could presumably be explained by the fact that the larger, more diverse dataset allows the neural network to learn underlying trends and characteristics in dose that are common to all treatment applicators. Accurate, applicator‐specific dose predictions could enable automated, knowledge‐based planning for any cervical brachytherapy treatment.

A novel skeletal muscle quantitative method and deep learning‐based sarcopenia diagnosis for cervical cancer patients treated with radiotherapy

AbstractBackgroundSarcopenia is associated with decreased survival in cervical cancer patients treated with radiotherapy. Cone‐beam computed tomography (CBCT) was widely used in image‐guided radiotherapy. Sarcopenia is assessed by the skeletal muscle index (SMI) of third lumbar vertebra (L3). Whereas, L3 is usually not included on the cervical cancer radiotherapy CBCT images.PurposeWe aimed to explore the usefulness of CBCT for evaluating SMI and deep learning (DL)‐based automatic segmentation and sarcopenia diagnosis for cervical cancer radiotherapy patients. We evaluated the SMI through fifth lumbar vertebra (L5).MethodsFirst, L3, L5 skeletal muscle area (SMA) were measured on CT and CBCT. The agreement of L5 skeletal muscle segmentation on CBCT was evaluated using the intraclass correlation coefficient (ICC). The relationships between L5‐SMICT and L3‐SMICT, L5‐SMICBCT were established and assessed by Pearson analysis, Bland‐Altman plots. Second, the consequent CBCT images of 248 cervical cancer radiotherapy patients with whole L5 were collected as DL‐based automatic segmentation. An independent external validation dataset was used. We proposed an end‐to‐end anatomical distance‐guided dual branch feature fusion network to segment L5 skeletal muscle on CBCT images. The automatic segmentation results were used for sarcopenia diagnosis evaluation.ResultsThe ICC values were greater than 0.95. The Pearson correlation coefficients (PCC) between L5‐SMICT and L3‐SMICT is 0.894. The PCC between L5‐SMICT and L5‐SMICBCT is 0.917. The L3‐SMICT could be estimated through L5‐SMICBCT by a linear regression equation. The adjusted R2 values were greater than 0.7. The dice similarity coefficient of automatic segmentation is 87.09%. Our proposed DL network predicted sarcopenia with 84.38% accuracy and 85.71% F1‐score. In external validation dataset, the sarcopenia diagnosis accuracy and F1‐score are 80% and 82.61%, respectively.ConclusionThe SMI quantitative measurement using CBCT for cervical cancer patients is feasible. And the DL network has the potential to assist in the sarcopenia diagnosis using CBCT images.

A workflow to select local tolerance limits by combining statistical process control and error curve model

Abstract Background Patient‐specific quality assurance (QA) is a complicated process specific to personnel, equipment, and procedure. The universal or commonly used tolerance limits may not be applicable to local situations. Therefore, it is a need for a medical physicist to establish appropriate local tolerance limits based on actual situations and quantitatively evaluate the error sensitivity of selected tolerance limits to determine their availability in clinical practice. Purpose This study aims to develop a comprehensive and scientifically sound methodology for determining appropriate local tolerance limits in patient‐specific QA. Methods and materials A total of 214 RapidArc plans for cervical cancer were selected. Systematic multi‐leaf collimator (MLC) positional errors were simulated across eighteen offsets ranging from ± 0.2 to ± 5 mm. Dose verification was conducted on 808 RapidArc plans, and a retrospective review was carried out. Firstly, six commonly used QA metrics in gamma and DVH analysis were extracted from the QA results of 196 error‐free RapidArc plans. These QA metrics included GP 10 (gamma passing rates [GPRs] at 3%/2mm, 10% dose threshold), GP 50 (GPRs at 3%/2mm, 50% dose threshold), µGI 50 (mean gamma index at 3%/2mm, 50% dose threshold), PTV 95 (dose received by 95% of PTV), PTV 5 (dose received by 5% of PTV) and PTV mean (mean dose received by PTV). Secondly, the statistical process control was used to establish the corresponding tolerance limits for each metric. Then, six error curve models were created based on 360 error‐introduced plans to record changes in QA metrics under different magnitudes of MLC positional error. The error range of theoretical detection limits for systematic MLC positional errors was investigated to assess error sensitivity quantitatively using the error curve model. Finally, the process‐based tolerance limits of six single QA metrics and four combined QA metrics were validated by using 252 sets of test data. The binary classification performance (error‐free/error‐introduced) was assessed based on detection rate, accuracy, precision, recall, and f1‐score. Results The theoretical detection limits for process‐based tolerance limits of GP 10 , GP 50 , µGI 50 , PTV 95 , PTV mean , and PTV 5 were 2.19 mm, 2.71 mm, 3.52 mm, 1.93 mm, 3.20 mm, and 2.15 mm, respectively. In the validation phase, the process‐based tolerance limits for PTV 95 effectively identified systematic MLC positional errors exceeding 0.6 mm with a detection rate of 76.19%, displaying superior performance in binary classification among six single metrics. Regarding combined metrics, the joint evaluation of process‐based tolerance limits for GP 10 and PTV 95 showed a higher detection rate of 80.16% for systematic MLC positional errors exceeding 0.6 mm. Conclusion The proposed workflow integrates the establishment and validation of tolerance limits. It not only provides a practical tool for setting local tolerance limits based on actual clinical scenarios but also offers a quantitative method for medical physicists to understand the error sensitivity of the selected local tolerance limits.

Transducer module apodization to reduce bone heating during focused ultrasound uterine fibroid ablation with phased arrays: A numerical study

AbstractBackgroundDuring magnetic resonance‐guided focused ultrasound (MRgFUS) surgery for uterine fibroids, ablation of fibrous tissues in proximity to the hips and spine is challenging due to heating within the bone that can cause patients to experience pain and potentially damage nerves. This far‐field bone heating limits the volume of fibroid tissue that is treatable via MRgFUS.PurposeTo investigate transducer module apodization for improving the ratio of focal‐to‐bone heating () when targeting fibroid tissue close to the hips and spine, to enable MRgFUS treatments closer to the bone.MethodsAcoustic and thermal simulations were performed using 3D magnetic resonance imaging (MRI)‐derived anatomies of ten patients who underwent MRgFUS ablation for uterine fibroids using a low‐frequency () 6144‐element flat fully‐populated modular phased array system (Arrayus Technologies Inc., Burlington, Canada) at our institution as part of a larger clinical trial (NCT03323905). Transducer modules ( per module) whose beams intersected with no‐pass zones delineated within the field were identified, their output power levels were reduced by varying blocking percentage levels, and the resulting temperature field distributions were evaluated across multiple sonications near the hip and spine bones in each patient. Acoustic and thermal simulations took approximately () and () to run for a single near‐spine (near‐hip) target, respectively.ResultsFor all simulated sonications, transducer module blocking improved compared to the no blocking case. In just over half of sonications, full module blocking maximized (increase of 82% 38% in 50% of hip targets and 49% 30% in 62% of spine targets vs. no blocking; mean ± SD), at the cost of more diffuse focusing (focal heating volumes increased by 13% ± 13% for hip targets and 39% ± 27% for spine targets) and thus requiring elevated total (hip: 6% ± 17%, spine: 37% ± 17%) and peak module‐wise (hip: 65% ± 36%, spine: 101% ± 56%) acoustic power levels to achieve equivalent focal heating as the no blocking control case. In the remaining sonications, partial module blocking provided further improvements in both (increased by 29% ± 25% in the hip and 15% ± 12% in the spine) and focal heating volume (decrease of 20% ± 10% in the hip and 34% ± 17% in the spine) relative to the full blocking case. The optimal blocking percentage value was dependent on the specific patient geometry and target location of interest. Although not all individual target locations saw the benefit, element‐wise phase aberration corrections improved the average compared to the no correction case (increase of 52% ± 47% in the hip, 35% ± 24% in the spine) and impacted the optimal blocking percentage value. Transducer module blocking enabled ablative treatments to be carried out closer to both hip and spine without overheating or damaging the bone (no blocking: /, full blocking: /, optimal partial blocking: / for hip/spine).ConclusionThe proposed transducer apodization scheme shows promise for improving MRgFUS treatments of uterine fibroids, and may ultimately increase the effective treatment envelope of MRgFUS surgery in the body by enabling tissue ablation closer to bony structures.

Individual curved‐needle interstitial template created using three‐dimensional printing for brachytherapy for distal parauterine tumor recurrence

AbstractBackgroundAchieving a clinically acceptable dose distribution with commercial vaginal applicators for brachytherapy of recurrent parauterine tumors is challenging. However, the application of three‐dimensional (3D) printing technology in brachytherapy has been widely acknowledged and can improve clinical treatment outcomes.PurposeThis study aimed to introduce an individual curved‐needle interstitial template (ICIT) created using 3D printing technology for high‐dose‐rate (HDR) brachytherapy with interstitial treatment to provide a clinically feasible approach to distal parauterine and vaginal cuff tumors. The entire workflow, including the design, optimization, and application, is presented.MethodsTen patients with pelvic cancer recurrence were examined at our center. The vaginal topography was filled with gauze strips soaked in developer solution, and images were obtained using computed tomography (CT) and magnetic resonance imaging (MRI). Curved needle paths were designed, and ICITs were 3D‐printed according to the high‐risk clinical target volume (HRCTV) and vaginal filling model. The dose and volume histogram parameters of the HRCTV (V100, V200, D90, and D98) and organs at risk (OARs) (D2cc) were recorded.ResultsAll patients completed interstitial brachytherapy treatment with the 3D‐printed ICIT. One patient experienced vaginal cuff tumor recurrence, and nine patients experienced parametrial tumor recurrence (four on the left and five on the right). We used two to five interstitial needles, and the maximum angle of the curved needle was 40°. No source obstruction events occurred during treatment of these 10 patients. The doses delivered to the targets and OARs of all patients were within the dose limits and based on clinical experience at our center.ConclusionThe ICIT is a treatment option for patients with distal parauterine tumor recurrence. This method addresses the limitations of vaginal intracavitary and standard interstitial applicators. The ICIT has the advantages of biocompatibility, personalization, and magnetic resonance imaging compatibility.

Uncertainty‐aware refinement framework for ovarian tumor segmentation in CECT volume

AbstractBackgroundOvarian cancer is a highly lethal gynecological disease. Accurate and automated segmentation of ovarian tumors in contrast‐enhanced computed tomography (CECT) images is crucial in the radiotherapy treatment of ovarian cancer, enabling radiologists to evaluate cancer progression and develop timely therapeutic plans. However, automatic ovarian tumor segmentation is challenging due to factors such as inhomogeneous background, ambiguous tumor boundaries, and imbalanced foreground‐background, all of which contribute to high predictive uncertainty for a segmentation model.PurposeTo tackle these challenges, we propose an uncertainty‐aware refinement framework that aims to estimate and refine regions with high predictive uncertainty for accurate ovarian tumor segmentation in CECT images.MethodsTo this end, we first employ an approximate Bayesian network to detect coarse regions of interest (ROIs) of both ovarian tumors and uncertain regions. These ROIs allow a subsequent segmentation network to narrow down the search area for tumors and prioritize uncertain regions, resulting in precise segmentation of ovarian tumors. Meanwhile, the framework integrates two guidance modules that learn two implicit functions capable of mapping query features sampled according to their uncertainty to organ or boundary manifolds, guiding the segmentation network to facilitate information encoding of uncertain regions.ResultsFirstly, 367 CECT images are collected from the same hospital for experiments. Dice score, Jaccard, Recall, Positive predictive value (PPV), 95% Hausdorff distance (HD95) and Average symmetric surface distance (ASSD) for the testing group of 77 cases are 86.31%, 73.93%, 83.95%, 86.03%, 15.17  mm and 2.57  mm, all of which are significantly better than that of the other state‐of‐the‐art models. And results of visual comparison shows that the compared methods have more mis‐segmentation than our method. Furthermore, our method achieves a Dice score that is at least 20% higher than the Dice scores of other compared methods when tumor volumes are less than 20 cm, indicating better recognition ability to small regions by our method. And then, 38 CECT images are collected from another hospital to form an external testing group. Our approach consistently outperform the compared methods significantly, with the external testing group exhibiting substantial improvements across key evaluation metrics: Dice score (83.74%), Jaccard (69.55%), Recall (82.12%), PPV (81.61%), HD95 (12.31 mm), and ASSD (2.32 mm), robustly establishing its superior performance.ConclusionsExperimental results demonstrate that the framework significantly outperforms the compared state‐of‐the‐art methods, with decreased under‐ or over‐segmentation and better small tumor identification. It has the potential for clinical application.

Tissue classification from raw diffusion‐weighted images using machine learning

Abstract Background In diffusion‐weighted imaging (DWI), a large collection of diffusion models is available to provide insights into tissue characteristics. However, these models are limited by predefined assumptions and computational challenges, potentially hindering the full extraction of information from the diffusion MR signal. Purpose This study aimed at developing a MOdel‐free Diffusion‐wEighted MRI (MODEM) method for tissue differentiation by using a machine learning (ML) algorithm based on raw diffusion images without relying on any specific diffusion model. MODEM has been applied to both simulation data and cervical cancer diffusion images and compared with several diffusion models. Methods With Institutional Review Board approval, 54 cervical cancer patients (median age, 52 years; age range, 29–73 years) participated in the study, including 26 in the early FIGO (International Federation of Gynecology and Obstetrics) stage (IB, 16; IIA, 10) and 28 the late stage (IIB, 8; IIIB, 14; IIIC, 1; IVA, 3; IVB, 2). The participants underwent DWI with 17 b ‐values (0 to 4500 s/mm 2 ) at 3 Tesla. Synthetic diffusion MRI signals were also generated using Monte‐Carlo simulation with Gaussian noise doping under varying substrates. MODEM with multilayer perceptron and five diffusion models (mono‐exponential, intra‐voxel incoherent‐motion, diffusion kurtosis imaging, fractional order calculus, and continuous‐time‐random‐walk models) were employed to distinguish different substrates in the simulation data and differentiate different pathological states (i.e., normal vs. cancerous tissue; and early‐stage vs. late‐stage cancers) in the cervical cancer dataset. Accuracy and area under the receiver operating characteristic (ROC) curve were evaluated. Mann–Whitney U‐test was used to compare the area under the curve (AUC) and accuracy values between MODEM and the five diffusion models. Results For the simulation dataset, MODEM produced a higher AUC and better accuracy, particularly in scenarios where the noise level exceeded 5%. For the cervical cancer dataset, MODEM yielded the highest AUC and accuracy in cervical cancer detection (AUC, 0.976; accuracy, 91.9%) and cervical cancer staging (AUC, 0.773; accuracy, 69.2%), significantly outperforming any of the diffusion models ( p  < 0.05). Conclusions MODEM is useful for cervical cancer detection and staging and offers considerable advantages over analytical diffusion models for tissue characterization.

Automatic dose prediction using deep learning and plan optimization with finite‐element control for intensity modulated radiation therapy

AbstractBackgroundAutomatic solutions for generating radiotherapy treatment plans using deep learning (DL) have been investigated by mimicking the voxel's dose. However, plan optimization using voxel‐dose features has not been extensively studied.PurposeThis study aims to investigate the efficiency of a direct optimization strategy with finite elements (FEs) after DL dose prediction for automatic intensity‐modulated radiation therapy (IMRT) treatment planning.MethodsA double‐UNet DL model was adapted for 220 cervical cancer patients (200 for training and 20 for testing), who underwent IMRT between 2016 and 2020 at our clinic. The model inputs were computed tomography (CT) slices, organs at risk (OARs), and planning target volumes (PTVs), and the outputs were dose distributions of uniformly generated high‐dose region‐controlled plans. The FEs were discretized into equal intervals of the dose prediction value within the [OARs avoid PTV(O‐P)] and [body avoids OARs & PTV(B‐OP)] regions in the test cohort and used to define the objectives for IMRT plan optimization. The plans were optimized using a two‐step process. In the beginning, the plans of two extra cases with and without low‐dose region control were compared to pursue robust and optimal dose adjustment degree pattern of FEs. In the first step, the mean dose of O‐P FEs were constrained to differing degrees according to the pattern. The further the FEs from the PTV, the tighter the constraints. In the second step, the mean dose of O‐P FEs from first step were constrained again but weakly and the dose of the B‐OP FEs from dose prediction and PTV were tightly regulated. The dosimetric parameters of the OARs and PTV were evaluated and compared using an interstep approach. In another 10 cases, the plans optimized via the aforementioned steps (method 1) were compared with those directly generated by the double‐UNet dose prediction model trained by low and high region‐controlled plans (method 2).ResultsThe mean differences in dose metrics between the UNet‐predicted dose and the clinical plans were: 0.47 Gy for bladder D50%; 0.62 Gy for rectum D50%; 0% for small intestine V30Gy; 1% for small intestine V40Gy; 4% for left femoral head V30Gy; and 6% for right femoral head V30Gy. The reductions in mean dose (p < 0.001) after FE‐based optimization were: 4.0, 1.9, 2.8, 5.9, and 5.7 Gy for the bladder, rectum, small intestine, left femoral head, and right femoral head, respectively, with flat PTV homogeneity and conformity. Method 1 plans produced lower mean doses than those of method 2 for the bladder (0.7 Gy), rectum (1.0 Gy), and small intestine (0.6 Gy), while maintaining  PTV homogeneity and conformity.ConclusionFE‐based direct optimization produced lower OAR doses and adequate PTV doses after DL prediction. This solution offers rapid and automatic plan optimization without manual adjustment, particularly in low‐dose regions.

A novel image segmentation network with multi‐scale and flow‐guided attention for early screening of vaginal intraepithelial neoplasia (VAIN)

Abstract Background Vaginal intraepithelial neoplasia (VAIN) is a rare precancerous lesion, and early diagnosis is crucial for preventing its progression to invasive vaginal cancer. However, the subtle differences in morphology and color between VAIN lesions and normal vaginal tissue make the automatic segmentation of VAIN highly challenging. Existing methods struggle to achieve precise segmentation, impacting the efficiency of early screening. Purpose This study aims to develop a high‐accuracy, robust deep learning image segmentation network to accurately and automatically segment VAIN lesions, thereby improving the efficiency and accuracy of early VAIN screening. Methods We propose a multi‐scale dilated attention flow network for VAIN image segmentation. This network improves upon the U‐Net architecture by optimizing the designs of the encoder and decoder and incorporating skip connection modules. In the encoding stage, we introduce the dilated squeeze‐and‐excitation (DiSE) module and the flow field guided adaptive separation and enhancement (FGASE) module. The DiSE module integrates dilated convolutions with varying dilation rates and a channel attention mechanism, effectively extracting multi‐scale contextual information and enhancing the model's ability to perceive VAIN lesions of different sizes. The FGASE module employs flow‐guided techniques to dynamically separate the features of the main region (VAIN lesions) from the edge region and enhance them individually. In the decoding stage, we propose a depth wise enhanced pooling (DEP) module that combines deep convolutional layers with adaptive pooling strategies to improve local feature extraction capabilities and optimize global contextual information. The skip connection stage introduces a triple statistical attention (TSA) module that utilizes global average pooling, global max pooling, and global standard deviation pooling to effectively capture diverse feature information, thereby enhancing the model's ability to model long‐range dependencies. Results Experiments conducted on a VAIN image dataset comprising 1142 patients demonstrate that the proposed network significantly outperforms other medical image segmentation methods across six metrics: Mean intersection over union (MIoU), dice coefficient, accuracy, recall, precision, and mean absolute error (MAE). Specifically, this network achieved an MIoU of 0.8461 and a Dice coefficient of 0.9166, substantially higher than other comparative methods, with a faster convergence speed. Ablation studies further confirm the effectiveness of each module in enhancing the model's performance. Conclusions The proposed network exhibits exceptional performance and robustness in the task of VAIN image segmentation, effectively segmenting VAIN lesions and providing strong technical support for early VAIN screening and clinical diagnosis. This work has significant clinical application value.

Impact of number of observers on reproducibility of gynaecological cancer MRI radiomics to interobserver contour variation

Abstract Background Current studies assessing the reproducibility of radiomic features from gynaecological magnetic resonance images (MRIs) with interobserver contour variation (IOV) have been limited to ≤3 observers. This number of observers is insufficient to demonstrate the full range of IOV. Purpose To assess the impact of observer numbers when investigating the reproducibility of gynaecological T2W‐MRI radiomic features with IOV. Methods 20 gynaecological cancer T2W‐MRIs had the gross tumor volume (GTV), bladder, rectum, uterus, parametrium, and vagina delineated by 6 observers to create a 2‐, 3‐, 4‐, 5‐, and 6‐observer dataset for each patient. IOV was assessed for each observer dataset and structure using the dice similarity coefficient, mean surface distance, and mean volume overlap variance. 107 radiomic features were extracted from each observer contour using PyRadiomics. The reproducibility of each radiomic feature was assessed for each observer dataset and structure using an intraclass correlation coefficient (ICC). An ICC estimate greater than 0.75 or 0.90 was classified as having good or excellent reproducibility, respectively. Results The GTV had a decrease in the number of features with good/excellent reproducibility when the number of observers in the dataset increased. Volumes with less IOV, such as the bladder and uterus, did not show this same trend, with consistent numbers of features with good/excellent reproducibility across all observer datasets. Conclusion Determining the reproducibility of gynaecological T2W‐MRI radiomic features to IOV with three or fewer observers is not adequate to display the full impact of IOV for GTVs.

On the impact of absorbed dose specification, tissue heterogeneities, and applicator heterogeneities on Monte Carlo‐based dosimetry of Ir‐192, Se‐75, and Yb‐169 in conventional and intensity‐modulated brachytherapy for the treatment of cervical cancer

PurposeThe purpose of this study was to evaluate the impact of dose reporting schemes and tissue/applicator heterogeneities for 192Ir‐, 75Se‐, and 169Yb‐based MRI‐guided conventional and intensity‐modulated brachytherapy.Methods and MaterialsTreatment plans using a variety of dose reporting and tissue/applicator segmentation schemes were generated for a cohort (n = 10) of cervical cancer patients treated with 192Ir‐based Venezia brachytherapy. Dose calculations were performed using RapidBrachyMCTPS, a Geant4‐based research Monte Carlo treatment planning system. Ultimately, five dose calculation scenarios were evaluated: (a) dose to water in water (Dw,w); (b) Dw,w taking the applicator material into consideration (Dw,wApp); (c) dose to water in medium (Dw,m); (d and e) dose to medium in medium with mass densities assigned either nominally per structure (Dm,m (Nom)) or voxel‐by‐voxel (Dm,m).ResultsIgnoring the plastic Venezia applicator (Dw,wApp) overestimates Dm,m by up to 1% (average) with high energy source (192Ir and 75Se) and up to 2% with 169Yb. Scoring dose to water (Dw,wApp or Dw,m) generally overestimates dose and this effect increases with decreasing photon energy. Reporting dose other than Dm,m (or Dm,m Nom) for 169Yb‐based conventional and intensity‐modulated brachytherapy leads to a simultaneous overestimation (up to 4%) of CTVHR D90 and underestimation (up to 2%) of bladder D2cc due to a significant dip in the mass‐energy absorption ratios at the depths of nearby targets and OARs. Using a nominal mass‐density assignment per structure, rather than a CT‐derived voxel‐by‐voxel assignment for MRI‐guided brachytherapy, amounts to a dose error up to 1% for all radionuclides considered.ConclusionsThe effects of the considered dose reporting schemes trend correspondingly between conventional and intensity‐modulated brachytherapy. In the absence of CT‐derived mass densities, MRI‐only‐based dosimetry can adequately approximate Dm,m by assigning nominal mass densities to structures. Tissue and applicator heterogeneities do not significantly impact dosimetry for 192Ir and 75Se, but do for 169Yb; dose reporting must be explicitly defined since Dw,m and Dw,w may overstate the dosimetric benefits.

An endovaginal MRI array with a forward‐looking coil for advanced gynecological cancer brachytherapy procedures: Design and initial results

AbstractPurposeTo develop an endovaginal MRI array that provides signal enhancement forward into the posterior parametrium and sideways into the vaginal wall, accelerating multiple‐contrast detection of residual tumors that survive external beam radiation. The array's enclosure should form an obturator for cervical cancer brachytherapy, allowing integration with MRI‐guided catheter placement, CT, and interstitial radiation dose delivery.MethodsThe endovaginal array consisted of forward‐looking and sideways‐looking components. The forward‐looking element imaged the cervix and posterior endometrium, and the sideways‐looking elements imaged the vaginal wall. Electromagnetic simulation was performed to optimize the geometry of a forward‐looking coil placed on a conductive‐metallic substrate, extending the forward penetration above the coil's tip. Thereafter, an endovaginal array with one forward‐looking coil and four sideways‐looking elements was constructed and tested at 1.5 Tesla in saline and gel phantoms, and three sexually mature swine. Each coil's tuning, matching, and decoupling were optimized theoretically, implemented with electronic circuits, and validated with network‐analyzer measurements. The array enclosure emulates a conventional brachytherapy obturator, allowing use of the internal imaging array together with tandem coils and interstitial catheters, as well as use of the enclosure alone during CT and radiation delivery. To evaluate the receive magnetic field () spatial profile, the endovaginal array's specific absorption‐rate (SAR) distribution was simulated inside a gel ASTM phantom to determine extreme heating locations in advance of a heating test. Heating tests were then performed during high SAR imaging in a gel phantom at the predetermined locations, testing compliance with MRI safety standards. To assess array imaging performance, signal‐to‐noise‐ratios (SNR) were calculated in a saline phantom and in vivo. Swine images were acquired with the endovaginal array combined with the scanner's body and spine arrays.ResultsSimulated profiles for the forward‐looking lobe pattern, obtained while varying several geometric parameters, disclosed that a forward‐looking coil placed on a metal‐backed substrate could double the effective forward penetration from approximately 25 to ∼40 mm. An endovaginal array, enclosed in an obturator enclosure was then constructed, with all coils tuned, matched, and decoupled. The ASTM gel‐phantom SAR test showed that peak local SAR was 1.2 W/kg in the forward‐looking coil and 0.3 W/kg in the sideways‐looking elements, well within ASTM/FDA/IEC guidelines. A 15‐min 4 W/kg average SAR imaging experiment resulted in less than 2oC temperature increase, also within ASTM/FDA/IEC heating limits. In a saline phantom, the forward‐looking coil and sideways‐looking array's SNR was four to eight times, over a 20–30 mm field‐of‐view (FOV), and five to eight times, over a 15–25 mm FOV, relative to the spine array's SNR, respectively. In three sexually mature swine, the forward‐looking coil provided a 5 + 0.2 SNR enhancement factor within the cervix and posterior endometrium, and the sideways‐looking array provided a 4 + 0.2 SNR gain factor in the vaginal wall, relative to the Siemens spine array, demonstrating that the array could significantly reduce imaging time.ConclusionsHigher SNR gynecological imaging is supported by forward‐looking and sideways‐looking coils. A forward‐looking endovaginal coil for cervix and parametrium imaging was built with optimized metal backing. Array placement within an obturator enhanced integration with the brachytherapy procedure and accelerated imaging for detecting postexternal‐beam residual tumors.

Radiobiological comparison between Cobalt‐60 and Iridium‐192 high‐dose‐rate brachytherapy sources: Part I—cervical cancer

AbstractPurposeThis study aimed to compare the biological effective doses (BEDs) to clinical target volume (CTV) and organs at risk (OARs) for cervical cancer patients treated with high‐dose‐rate (HDR) Iridium‐192 (192Ir) or Cobalt‐60 (60Co) brachytherapy (BT) boost and to determine if the radiobiological differences between the two isotopes are clinically relevant.MethodsConsidering all radiosensitivity parameters and their reported variations, the BEDs to CTV and OARs during HDR 60Co/192Ir BT boost were evaluated at the voxel level. The anatomical differences between individuals were also taken into account by retrospectively considering 25 cervical cancer patients. The intrafraction repair, proliferation, hypoxia‐induced radiosensitivity heterogeneity, relative biological effectiveness (RBE), and source aging dose‐rate variation were also taken into account. The comparisons in CTV were performed based on equivalent uniform BED (EUBED).ResultsConsidering nominal parameters with no RBE correction, the CTV EUBEDs were almost similar with a median ratio of ∼1.00 (p < 0.00001), whereas RBE correction resulted in 3.9%–5.5% (p = 0.005, median = 4.8%) decrease for 60Co with respect to 192Ir. For OARs, the median values of D2cc (in EQD23) for 60Co were lower than that of 192Ir up to 9.2% and 11.3% (p < 0.00001) for nominal parameters and fast repair conditions, respectively. In addition, for a nominal value (reported range) of radiosensitive parameters, the CTV EUBED differences of up to 6% (5%–10%) were assessed for HDR‐BT component.ConclusionThe RBE values are the most important cause of discrepancies between the two sources. By comparing BED/EUBEDs to CTV and OARs between 60Co and 192Ir sources, this numerical study suggests that a dose escalation to ∼4% is feasible and safe while sparing well the surrounding normal tissues. This 4% dose escalation should be benchmarked with clinical evidences (such as the results of clinical trials) before it can be used in clinical practice.

Clinical acceptability of fully automated external beam radiotherapy for cervical cancer with three different beam delivery techniques

AbstractPurposeTo fully automate CT‐based cervical cancer radiotherapy by automating contouring and planning for three different treatment techniques.MethodsWe automated three different radiotherapy planning techniques for locally advanced cervical cancer: 2D 4‐field‐box (4‐field‐box), 3D conformal radiotherapy (3D‐CRT), and volumetric modulated arc therapy (VMAT). These auto‐planning algorithms were combined with a previously developed auto‐contouring system. To improve the quality of the 4‐field‐box and 3D‐CRT plans, we used an in‐house, field‐in‐field (FIF) automation program. Thirty‐five plans were generated for each technique on CT scans from multiple institutions and evaluated by five experienced radiation oncologists from three different countries. Every plan was reviewed by two of the five radiation oncologists and scored using a 5‐point Likert scale.ResultsOverall, 87%, 99%, and 94% of the automatically generated plans were found to be clinically acceptable without modification for the 4‐field‐box, 3D‐CRT, and VMAT plans, respectively. Some customizations of the FIF configuration were necessary on the basis of radiation oncologist preference. Additionally, in some cases, it was necessary to renormalize the plan after it was generated to satisfy radiation oncologist preference.ConclusionApproximately, 90% of the automatically generated plans were clinically acceptable for all three planning techniques. This fully automated planning system has been implemented into the radiation planning assistant for further testing in resource‐constrained radiotherapy departments in low‐ and middle‐income countries.

The versatility of evolutionary intelligent tri‐objective treatment planning for cervical cancer brachytherapy

Abstract Background A multi‐objective automated treatment planning approach, called BRIGHT, has demonstrated success in prostate cancer brachytherapy (BT). BRIGHT optimizes directly on dose‐volume metrics, aligning with clinical protocol goals, and produces multiple plans that represent different trade‐offs between tumor coverage and healthy organ sparing. Current automated treatment planning methods either do not optimize directly on dose‐volume metrics or generate a single plan, which is only considered optimal in the specific optimization model. Purpose We extended BRIGHT to cervical cancer BT, for which adding a third objective to the existing bi‐objective approach was deemed necessary. In this work, we present the algorithmic adaptations made to the approach and highlight its flexibility, which enables straightforward inclusion of customizations. We further demonstrate that this approach produces clinically acceptable plans. Methods The first two objectives in the proposed approach pertain to the EMBRACE‐II protocol, which is divided into tumor coverage and healthy organ sparing. The third objective encompasses added aims, which were deemed necessary to be included to ensure dose distribution shape characteristics not captured in the EMBRACE‐II protocol but which can also readily be tuned to include local clinical preferences. We illustrate this by proposing four different customizations: a baseline customization and three different customizations that lead to (potentially distinct) pear‐shaped dose distributions, often desired in cervical cancer BT. We include optimization with contiguous volumes, a capability distinctive to BRIGHT, as an option for dose distribution shape optimization. We tested all four customizations on 269 BT fractions (123 patients), and studied differences in runtimes, 3D dose distributions, as well as obtained dose‐volume values. Clinical acceptability was evaluated for six representative patient cases, by presenting the resulting set of plans for all customizations to a BT team of two radiation oncologists, a medical physicist, and a radiation therapy technologist. They were asked to assess whether there is at least one acceptable plan per patient in the given set of plans. Results Treatment plans can be generated in under 2.8 min with the baseline tri‐objective BRIGHT, or 3.7 min if contiguous volumes are included, even though 260.000 dose calculation points are used for highly accurate dose estimation during optimization. There are visual differences in dose distributions for some of the six patient cases when using the distinct customizations, although generally pear‐shaped distributions were obtained. The contiguity of the dose distributions resulting from optimizing with contiguous volumes can be advantageous in special cases where the high‐dose region is preferred in the target area, as well as directly being tied to the location of the inserted applicator. Achieved dose‐volume values are clinically comparable between all four customizations. The BT team indicated that 3/4 customizations included at least one clinically acceptable plan for all six patients. Conclusions Clinically acceptable plans for cervical cancer BT can be quickly generated using the new tri‐objective version of BRIGHT. This approach allows for straightforward customization to accommodate local clinical preferences. We demonstrated this versatility through various customizations that produced generally pear‐shaped, yet potentially distinct, dose distributions, with comparable dose‐volume values according to the EMBRACE‐II protocol.

Accelerate treatment planning process using deep learning generated fluence maps for cervical cancer radiation therapy

AbstractPurposeThis study aims to develop a deep learning method that skips the time‐consuming inverse optimization process for automatic generation of machine‐deliverable intensity‐modulated radiation therapy (IMRT) plans.MethodsNinety cervical cancer clinical IMRT plans were collected to train a two‐stage convolution neural network, of which 66 plans were assigned for training, 11 for validation, and 13 for test. The neural network took patients’ computed tomography (CT) anatomy as the input and predicted the fluence map for each radiation beam. The predicted fluence maps were then imported into a treatment planning system and converted to multileaf collimators motion sequences. The automatic plan was evaluated against its corresponding clinical plan, and its machine deliverability was validated by patient‐specific IMRT quality assurance (QA).ResultsThere were no significant differences in dose parameters between automatic and clinical plans for all 13 test patients, indicating a good prediction of fluence maps and a decent quality of automatic plans. The average dice similarity coefficient of isodose volumes encompassed by 0%–100% isodose lines ranged from 0.94 to 1. In patient‐specific IMRT QA, the mean gamma passing rate of automatic plans achieved 99.5% under 3%/3 mm criteria, and 97.3% under 2%/2 mm criteria, with a low dose threshold of 10%.ConclusionsThe proposed deep learning framework can produce machine‐deliverable IMRT plans with quality similar to the clinical plans in the test set. It skips the inverse plan optimization process and provides an effective and efficient method to accelerate treatment planning process.

Automatic segmentation of magnetic resonance images for high‐dose‐rate cervical cancer brachytherapy using deep learning

AbstractPurposeMagnetic resonance (MR) imaging is the gold standard in image‐guided brachytherapy (IGBT) due to its superior soft‐tissue contrast for target and organs‐at‐risk (OARs) delineation. Accurate and fast segmentation of MR images are very important for high‐quality IGBT treatment planning. The purpose of this work is to implement and evaluate deep learning (DL) models for the automatic segmentation of targets and OARs in MR image‐based high‐dose‐rate (HDR) brachytherapy for cervical cancer.MethodsA 2D DL model using residual neural network architecture (ResNet50) was developed to contour the targets (gross tumor volume (GTV), high‐risk clinical target volume (HR CTV), and intermediate‐risk clinical target volume (IR CTV)) and OARs (bladder, rectum, sigmoid, and small intestine) automatically on axial MR slices of HDR brachytherapy patients. Furthermore, two additional 2D DL models using sagittal and coronal images were also developed. A 2.5D model was generated by combining the outputs from axial, sagittal, and coronal DL models. Similarly, a 2D and 2.5D DL models were also generated for the inception residual neural network (InceptionResNetv2 (InRN)) architecture. The geometric (Dice similarity coefficient (DSCs) and 95th percentile of Hausdorff distance (HD)) and dosimetric accuracy of 2D (axial only) and 2.5D (axial + sagittal + coronal) DL model generated contours were calculated and compared.ResultsThe mean (range) DSCs of ResNet50 across all contours were 0.674 (0.05–0.96) and 0.715 (0.26–0.96) for the 2D and 2.5D models, respectively. For InRN, these were 0.676 (0.11–0.96) and 0.723 (0.35–0.97) for the 2D and 2.5D models, respectively. The mean HD of ResNet50 across all contours was 15.6 mm (1.8–69 mm) and 12.1 mm (1.7–44 mm) for the 2D and 2.5D models, respectively. The similar results for InRN were 15.4 mm (2–68 mm) and 10.3 mm (2.7–39 mm) for the 2D and 2.5D models, respectively. The dosimetric parameters (D90) of GTV and HR CTV for manually contoured plans matched better with the 2.5D model (p > 0.6) and the results from the 2D model were slightly lower (p < 0.08). On the other hand, the IR CTV doses (D90) for all of the models were slightly lower (2D: ‐1.3 to ‐1.5 Gy and 2.5D: ‐0.5 to ‐0.6 Gy) and the differences were statistically significant for the 2D model (2D: p < 0.000002 and 2.5D: p > 0.06). In case of OARs, the 2.5D model segmentations resulted in closer dosimetry than 2D models (2D: p = 0.07–0.91 and 2.5D: p = 0.16–1.0).ConclusionsThe 2.5D DL models outperformed their respective 2D models for the automatic contouring of targets and OARs in MR image‐based HDR brachytherapy for cervical cancer. The InceptionResNetv2 model performed slightly better than ResNet50.

Improving predictive CTV segmentation on CT and CBCT for cervical cancer by diffeomorphic registration of a prior

AbstractPurposeAutomatic cervix‐uterus segmentation of the clinical target volume (CTV) on CT and cone‐beam CT (CBCT) scans is challenged by the limited visibility and the non‐anatomical definition of certain border regions. We study the potential performance gain of convolutional neural networks by regulating the segmentation predictions as diffeomorphic deformations of a segmentation prior.Materials and MethodsWe introduce a 3D convolutional neural network that segments the target scan by joint voxel‐wise classification and the registration of a given prior. We compare this network to two other 3D baseline models: One treating segmentation as a classification problem (segmentation‐only), the other as a registration problem (deformation‐only). For reference and to highlight the benefits of a 3D model, these models are also benchmarked against a 2D segmentation model. Network performances are reported for CT and CBCT segmentation of the cervix‐uterus CTV. We train the networks on the data of 84 patients. The prior is provided by the CTV segmentation of a planning CT. Repeat CT or CBCT scans constitute the target scans to be segmented.ResultsAll 3D models outperformed the 2D segmentation model. For CT segmentation, combining classification and registration in the proposed joint model proved beneficial, achieving a Dice score of 0.87 and a mean squared error (MSE) of the surface distance below 1.7 mm. No such synergy was observed for CBCT segmentation, for which the joint and the deformation‐only model performed similarly, achieving a Dice score of about 0.80 and an MSE surface distance of 2.5 mm. However, the segmentation‐only model performed notably worse in this low contrast regime. Visual inspection revealed that this performance drop translated into geometric inconsistencies between the prior and target segmentation. Such inconsistencies were not observed for the deformation‐based models.ConclusionConstraining the solution space of admissible segmentation predictions to those reachable by a diffeomorphic deformation of the prior proved beneficial as it improved geometric consistency. Especially for CBCT, with its poor soft‐tissue contrast, this type of regularization becomes important as shown by quantitative and qualitative evaluation.

Automatic clinical target volume delineation for cervical cancer in CT images using deep learning

PurposeAccurately delineating clinical target volumes (CTV) is essential for completing radiotherapy plans but is time‐consuming, labor‐intensive, and prone to inter‐observer variation. Automating CTV delineation has the benefits of both speeding up contouring process and improving the quality of contours. Recently, auto‐segmentation approaches based on deep learning have achieved some improvements. However, unlike organ segmentation, the CTV contains potential tumor spread tissues or subclinical disease tissues, resulting in poorly defined margin interface and irregular shape. It is not reasonable to directly apply the deep learning segmentation algorithms to CTV tasks without considering the unique characteristics of shape and margin. In this work, we propose a novel automatic CTV delineation algorithm based on deep learning addressing the unique shape and margin challenges.MethodsOur deep learning method, called RA‐CTVNet, segments the CTV from cervical cancer CT images. RA‐CTVNet denotes our automatic CTV delineation algorithm based on deep learning with Area‐aware reweight strategy and Recursive refinement strategy. (1) In order to process the whole‐volume CT images and delineate all CTVs in one shot, our method is built upon the popular 3D Unet architecture. We further extend it with robust residual learning and squeeze‐and‐excitation blocks for better feature representation. (2) We propose area‐aware reweight strategy which assigns different weights for different slices. The core is adjusting model’s attention to each slice. (3) In terms of the trade‐off between providing performance improvements and meeting the limitations of GPU memory, we exploit a new recursive refinement strategy to address margin challenge.ResultsThis retrospective study included 462 patients diagnosed with cervical cancer who received radiotherapy from June 2017 to May 2019. Extensive experiments were conducted to evaluate performance of RA‐CTVNet. First, compared to different network architectures, RA‐CTVNet achieved improvements in Dice similarity coefficient (DSC). Second, we conducted ablation study. The results showed that compared to the backbone, area‐aware reweight strategy increased DSC by 3.3% on average and recursive refinement strategy further increased DSC by 1.6% on average. Then, we compared our method with three human experts. Our RA‐CTVNet performed better than two experts while comparably to the third expert. Finally, a multicenter evaluation was conducted to verify the accuracy and generalizability.ConclusionsOur findings show that deep learning is able to offer an efficient framework for automatic CTV delineation. The tailored RA‐CTVNet can improve the quality of CTV contours, which has great potential for reducing the burden of experts and increasing the accuracy of delineation. In the future, if with more training data, further improvements are possible, bringing this approach closer to real clinical practice.

Geometrically focused training and evaluation of organs‐at‐risk segmentation via deep learning

Abstract Background Deep learning methods are promising in automating segmentation of organs at risk (OARs) in radiotherapy. However, the lack of a geometric indicator for dosimetry accuracy remains to be a problem. This issue is particularly pronounced in specific radiotherapy treatments where only the proximity of structures to the radiotherapy target affects the dose planning. In cervical cancer high dose‐rate (HDR) brachytherapy, treatment planning is motivated by limiting dose to the hottest 2 cubic centimeters (D2cm 3 ) of the OARs. Similarly, Ethos online adaptive radiotherapy system prioritizes only the closest target structures for adaptive plan generation. Purpose We propose a novel geometrically focused deep learning training method and evaluation metric, using cervical brachytherapy as a case study. A distance‐penalized (DP) loss function was developed to focus attention on the near‐to‐target OAR regions. We also introduced and evaluated a novel geometric metric, weighted dice similarity coefficient (wDSC), correlated with OARs D2cm 3 . Methods A model was trained using a 3D U‐Net architecture and 170 T2‐weighted magnetic resonance (MR) images (56 patients) with clinical contours. The dataset was split into subsets at the patient level: 45 patients (150 scans) as the training set for five‐fold cross‐validation and 11 patients (20 scans) as the testing set. Another dataset from our institution, consisting of 35 MR scans from 22 cervical cancer patients, was used as an independent internal testing set. A distance map, emphasizing errors near high‐risk clinical target volume (CTV HR ), was used to penalize two commonly used loss functions, cross‐entropy (CE) loss and DiceCE loss. The wDSC emphasizes the accuracy of OAR regions proximal to CTV HR by incorporating a weighted factor in the original vDSC. The Pearson correlation coefficient ( r ) was used to quantify the strength of the relationship between D2cm 3 accuracy and six evaluation metrics (wDSC and five standard metrics). A physician rated and revised the auto‐contours for the clinical acceptability tests. Results The wDSC moderately correlated ( r  = ‐0.55) with D2cm 3 accuracy, outperforming standard geometric metrics. Models using DP loss functions consistently yielded higher wDSCs compared to their respective non‐DP counterparts. DP loss models also improved D2cm 3 accuracy, indicating an enhanced accuracy in dosimetry. The clinical acceptability tests revealed that more than 94% of bladder and rectum contours and approximately half of the sigmoid and small bowel contours were clinically accepted. Conclusion We developed and evaluated a new geometric metric, wDSC, as a better indicator of D2cm 3 accuracy, which has the potential to become a surrogate for dosimetric accuracy in cervical brachytherapy. The model with DP loss showed non‐statistically significant improvements in geometric and dosimetric performance. This work also holds the potential to be used for precise OARs delineation in adaptive radiotherapy.

Megavoltage CT enhancement for cervical cancer tomotherapy using a generative adversarial network with deformable convolution and self‐attention

Abstract Background Megavoltage computed tomography (MVCT) is an essential imaging modality for verifying patient positioning in helical tomotherapy. However, its clinical application in daily anatomical monitoring and adaptive radiotherapy is hindered by inherent image artifacts and poor soft‐tissue contrast. This issue is particularly pronounced in pelvic radiotherapy, where the intra‐ and interfraction anatomical variations necessitate high‐quality image guidance to ensure precise dose delivery. Purpose We developed a deep‐learning‐based framework for MVCT enhancement to improve anatomical visualization and facilitate accurate adaptive treatment planning for cervical cancer. Methods This study analyzed a retrospective cohort of 170 patients with cervical cancer who underwent helical tomotherapy. The proposed deep‐learning‐based algorithm employed a generative adversarial network (GAN) that integrated deformable convolution and a self‐attention mechanism (SADC‐EGAN) to improve MVCT image quality. Comparative analyses were conducted against representative baseline methods, including U‐Net, Attention U‐Net, U‐Net++, Swin‐UNet, CycleGAN, and Pix2Pix. Model performance was assessed using quantitative metrics, including mean absolute error (MAE), peak signal‐to‐noise ratio (PSNR), structural similarity index measure (SSIM), and Fréchet inception distance (FID). Results The synthetic computed tomography (sCT) images generated by the proposed SADC‐EGAN method demonstrated superior Hounsfield unit (HU) accuracy and structural similarity compared to the original MVCT. Specifically, the MAE between the sCT and kilovoltage computed tomography (kVCT) was reduced to 36.32 ± 6.69 HU, compared with 56.72 ± 9.09 HU for MVCT. In terms of image quality, the sCT images exhibited notable enhancements over MVCT images, with higher PSNR (32.54 ± 2.31 vs. 29.40 ± 1.56 dB), improved SSIM (0.93 ± 0.01 vs. 0.89 ± 0.02), and substantially lower FID (66.68 ± 22.31 vs. 153.52 ± 28.77). Conclusions The proposed SADC‐EGAN framework, integrating deformable convolutions and self‐attention, effectively generated high‐quality kVCT‐like images from MVCT, improving both HU accuracy and image quality. This approach has clinical potential to enable online adaptive helical tomotherapy for cervical cancer.

Evaluating microstructures in endometrial cancer using diffusion‐relaxation correlated spectroscopic imaging: Histopathological correlations

Abstract Background Endometrial cancer (EC) is a prevalent gynecologic malignancy where accurate grading and assessment are crucial for determining prognosis and treatment strategies. Conventional MRI techniques, including apparent diffusion coefficient (ADC) and T2‐weighted imaging, often fail to capture the detailed microstructural complexities of EC. Purpose To evaluate the efficacy of diffusion relaxation correlated spectroscopic imaging (DR‐CSI) in assessing EC and to compare its diagnostic performance with conventional ADC and T2‐weighted imaging. Materials and Methods Sixty‐two patients with histopathologically confirmed EC were included in this prospective study. All patients underwent preoperative MRI, including DR‐CSI using a multi‐TE (50–90 ms) and multi‐b‐value (0–1600 s/mm 2 ) echo‐planar imaging sequence. The DR‐CSI data were analyzed to generate a four‐compartment D‐T2 spectra, yielding corresponding volume fraction metrics (VF, I–IV). Voxel‐wise ADC and T2 values were also obtained. The relationships between these imaging parameters and histopathologic results were evaluated using one‐way ANOVA or Kruskal–Wallis tests. Diagnostic performance was assessed using receiver operating characteristic (ROC) curve analysis. Results VF II and VF III demonstrated significant differences across histological grades ( p  < 0.01 and p  = 0.04, respectively). The combination of VF II and VF III provided optimal differentiation between low‐ and high‐grade EC (Area under curve, AUC 0.801 [95% confidence interval: 0.623–0.937]). VF IV exhibited superior performance in distinguishing lymph node metastasis (LNM) status (AUC 0.734 [0.556–0.892]). The combination of VF IV and VF II improved performance in predicting LNM status (AUC 0.826 [0.66–0.961]). However, no parameter alone effectively distinguished myometrial invasion (MI) statuses, but the combination of VF I and ADC improved performance (AUC 0.706 [0.560–0.844]). Conclusion DR‐CSI offers a novel and effective method for quantifying microstructural compartments in EC, providing superior diagnostic accuracy compared to conventional ADC and T2 values. The ability to capture detailed microstructural information from DR‐CSI metrics holds promise for improving EC diagnosis and grading, offering deeper insights into tumor heterogeneity.

Enhancing auto‐contouring with large language model in high‐dose rate brachytherapy for cervical cancers

Abstract BACKGROUND High‐dose‐rate brachytherapy (HDR‐BT) is a cornerstone of cervical cancer (CC) treatment, requiring the precise delineation of high‐risk clinical target volumes (HR‐CTV) and organs at risk (OARs) for effective dose delivery and toxicity reduction. However, the time‐sensitive nature of HDR‐BT planning and its reliance on expert contouring introduce inter‐ and intra‐observer variability, posing challenges for consistent and accurate treatment planning. PURPOSE This study proposes a novel deep learning (DL)‐based auto‐segmentation framework, guided by task‐specific prompts generated from large language models (LLMs), to address these challenges and improve segmentation accuracy and efficiency. METHODS A retrospective dataset of 32 CC patients, encompassing 124 planning computed tomography (pCT) images, was utilized. The framework integrates clinical guidelines for organ contouring from the American Brachytherapy Society (ABS), the European Society for Radiotherapy and Oncology (ESTRO), and the International Commission on Radiation Units and Measurements (ICRU). LLMs, particularly Chat‐GPT, extracts domain knowledge from these contouring guidelines to generate task‐specific prompts, which guide a Swin transformer‐based encoder and a fully convolutional network (FCN) decoder for segmentation. The DL pipeline was evaluated on HR‐CTV and OARs, including the bladder, rectum, and sigmoid. Metrics such as Dice similarity coefficient (DSC), Hausdorff distance (HD95%), mean surface distance (MSD), and center‐of‐mass distance (CMD) were used for performance assessment. An ablation study compared the prompt‐guided approach with a baseline model without prompt guidance. Statistical differences were tested with two‐tailed paired t ‐tests, and p ‐values were adjusted using the Benjamini–Hochberg method to address the multiple comparisons correction and results with adjusted p  < 0.05 were deemed significant. Cohen's d values were calculated to quantify effect sizes. RESULTS The proposed framework achieved the highest segmentation for the bladder (DSC of 0.91 ± 0.07), followed by the HR‐CTV (DSC of 0.80 ± 0.08) and the rectum (DSC of 0.78 ± 0.07), and a lower accuracy for sigmoid (DSC of 0.63 ± 0.15) due to its small size and irregular shape. Boundary precision was highest for the HR‐CTV (HD95%: 6.32 ± 2.31 mm). The ablation study confirmed the contribution of prompt guidance, with statistically significant improvements in DSC and/or HD95% ( p  < 0.05) for all OARs. Prompt guidance, however, did not improve the accuracy of HR‐CTV delineation. CONCLUSIONS This study demonstrates the feasibility and effectiveness of integrating LLM‐generated task‐specific prompts with DL‐based segmentation for HDR‐BT in CC. The proposed framework enhances segmentation consistency to support accurate treatment planning, addressing critical challenges in HDR‐BT workflows.

Multibody dynamic modeling of the behavior of flexible instruments used in cervical cancer brachytherapy

AbstractBackgroundThe steep radiation dose gradients in cervical cancer brachytherapy (BT) necessitate a thorough understanding of the behavior of afterloader source cables or needles in the curved channels of (patient‐tailored) applicators.PurposeThe purpose of this study is to develop and validate computer models to simulate: (1) BT source positions, and (2) insertion forces of needles in curved applicator channels. The methodology presented can be used to improve the knowledge of instrument behavior in current applicators and aid the development of novel (3D‐printed) BT applicators.MethodsFor the computer models, BT instruments were discretized in finite elements. Simulations were performed in SPACAR by formulating nodal contact force and motion input models and specifying the instruments’ kinematic and dynamic properties. To evaluate the source cable model, simulated source paths in ring applicators were compared with manufacturer‐measured source paths. The impact of discrepancies on the dosimetry was estimated for standard plans. To validate needle models, simulated needle insertion forces in curved channels with varying curvature, torsion, and clearance, were compared with force measurements in dedicated 3D‐printed templates.ResultsComparison of simulated with manufacturer‐measured source positions showed 0.5–1.2 mm median and <2.0 mm maximum differences, in all but one applicator geometry. The resulting maximum relative dose differences at the lateral surface and at 5 mm depth were 5.5% and 4.7%, respectively. Simulated insertion forces for BT needles in curved channels accurately resembled the forces experimentally obtained by including experimental uncertainties in the simulation.ConclusionThe models developed can accurately predict source positions and insertion forces in BT applicators. Insights from these models can aid novel applicator design with improved motion and force transmission of BT instruments, and contribute to the estimation of overall treatment precision. The methodology presented can be extended to study other applicator geometries, flexible instruments, and afterloading systems.

Dosimetric predictors and Lyman normal tissue complication probability model of hematological toxicity in cervical cancer patients with treated with pelvic irradiation

AbstractPurposeTo identify dosimetric parameters associated with acute hematological toxicity (HT) and identify the corresponding normal tissue complication probability (NTCP) model in cervical cancer patients receiving helical tomotherapy (Tomo) or fixed‐field intensity‐modulated radiation therapy (ff‐IMRT) in combination with chemotherapy, that is, concurrent chemoradiotherapy (CCRT) using the Lyman–Kutcher–Burman normal tissue complication probability (LKB‐NTCP) model.MethodsData were collected from 232 cervical cancer patients who received Tomo or ff‐IMRT from 2015 to 2018. The pelvic bone marrow (PBM) (including the ilium, pubes, ischia, acetabula, proximal femora, and lumbosacral spine) was contoured from the superior boundary (usually the lumbar 5 vertebra) of the planning target volume (PTV) to the proximal end of the femoral head (the lower edge of the ischial tubercle). The parameters of the LKB model predicting ≥grade 2 hematological toxicity (Radiation Therapy Oncology Group [RTOG] grading criteria) (TD50(1), m, and n) were determined using maximum likelihood analyses. Univariate and multivariate logistic regression analyses were used to identify correlations between dose–volume parameters and the clinical factors of HT.ResultsIn total, 212 (91.37%) patients experienced ≥grade 2 hematological toxicity. The fitted normal tissue complication probability model parameters were TD50(1) = 38.90 Gy (95%CI, [36.94, 40.96]), m = 0.13 (95%CI [0.12, 0.16]), and n = 0.04 (95%CI [0.02, 0.05]). Per the univariate analysis, the NTCP (the use of LKB‐NTCP with the set of model parameters found, p = 0.023), maximal PBM dose (p = 0.01), mean PBM dose (p = 0.021), radiation dose (p = 0.001), and V16–53 (p < 0. 05) were associated with ≥grade 2 HT. The NTCP (the use of LKB‐NTCP with the set of model parameters found, p = 0.023; AUC = 0.87), V16, V17, and V18 ≥ 79.65%, 75.68%, and 72.65%, respectively (p < 0.01, AUC = 0.66∼0.68), V35 and V36 ≥ 30.35% and 28.56%, respectively (p < 0.05; AUC = 0.71), and V47 ≥ 13.43% (p = 0.045; AUC = 0.80) were significant predictors of ≥grade 2 hematological toxicity from the multivariate logistic regression analysis.ConclusionsThe volume of the PBM of patients treated with concurrent chemoradiotherapy and subjected to both low‐dose (V16–18) and high‐dose (V35,36 and V47) irradiation was associated with hematological toxicity, depending on the fractional volumes receiving the variable degree of dosage. The NTCP were stronger predictors of toxicity than V16–18, V35, 36, and V47. Hence, avoiding radiation hot spots on the PBM could reduce the incidence of severe HT.

Endometrial cancer tissue features clusterization by kurtosis MRI

AbstractBackgroundEndometrial cancer (EC) is one of the most common gynecological malignancies and the second most common gynecological malignancy cause of death in women. Heterogeneous tissues with different grades of complexity and different diffusion properties characterize the EC. Several diffusion magnetic resonance imaging (DMRI) protocols have been used to perform a non‐invasive and global evaluation of EC for diagnostic and prognostic purposes. However, the association of a single value for the diffusion coefficient to an EC tissue could be a severe limit for developing a DMRI virtual histology protocol.PurposeThis study evaluates the potential of diffusion kurtosis imaging (DKI) and tissue multiple diffusion clusterization in detecting the specific features of healthy/cancer tissue that can be useful in EC diagnosis and prognosis.MethodsThirty‐eight subjects were analyzed: 18 with a final diagnosis of EC and 20 healthy, asymptomatic, with no history of endometrial pathology and uterine tumor pathology. Diffusion‐weighted Spin‐Echo Echo‐Planar Imaging (DW‐EPI) with TR/TE = 2000 ms/77 ms was used at 3T using six different b‐values: (500, 800, 1000, 1500, 2000, and 2500)s/mm2 along three gradient directions (x, y, z). The decay of the signal in each voxel was used to obtain clusters of different diffusion compartments reflecting tissue heterogeneity. Moreover, using the Kurtosis representation, the parametric maps of the apparent kurtosis (K) and diffusivity (D) coefficients were obtained. The statistical analysis of the differences in the mean value of the parameters obtained in the selected regions of interest (ROIs) in tumor area (T) peritumor area (PT) and healthy tissue was carried out using a Kruskal–Wallis Test. A p‐value < 0.05 indicated a statistically significant difference. To validate DKI and multiple diffusion clusterization in the detection of EC and healthy tissue, DMRI results were compared with EC histology. A ROC curve analysis was performed to evaluate the performance of the clustering feature in differentiating healthy and tumoral tissues.ResultsK discriminates the peritumor area (PT) of the tumor from the healthy tissues (p < 0.05) and the area inside the EC (cancerous tissue, p < 0.05). This result is validated and explained by the diffusion clustering, which shows a great variability in K for pathological compared to healthy subjects. Moreover, the standard deviation of K in the cluster defined by the highest K/D ratio differentiates T and H ROIs.ConclusionsK as well as diffusion clusterization are sensitive to the different microstructural organizations in EC and healthy tissue, promoting themself as a potential tool for the diagnosis and prognosis of EC.

Clinical test cases for commissioning, QA, and benchmarking of model‐based dose calculation algorithms in 1 ⁹ 2 Ir HDR gynecologic tandem and ring brachytherapy

Abstract Purpose To develop clinically relevant test cases for commissioning Model‐Based Dose Calculation Algorithms (MBDCAs) for 192 Ir High Dose Rate (HDR) gynecologic brachytherapy following the workflow proposed by the TG‐186 report and the WGDCAB report 372. Acquisition and validation methods Two cervical cancer intracavitary HDR brachytherapy models were developed based on a real patient, using either uniformly structured regions or realistic segmentation. The patient's computed tomography (CT) images were processed, converted to a series of digital imaging and communications in medicine (DICOM) CT images, and imported into two treatment planning systems (TPSs), the Oncentra Brachy and BrachyVision. The original segmentation of the clinical case was augmented to enable a thorough dosimetric analysis. The actual clinical treatment plan was generally maintained, with the source replaced by a generic 192 Ir HDR source. Dose to medium in medium calculations were performed using the MBDCA option of each TPS, and three different Monte Carlo (MC) simulation codes. MC results demonstrated agreement within statistical uncertainty, while comparisons between the commercial TPS MBDCAs and a general‐purpose MC code highlighted both the advantages and limitations of the studied MBDCAs, suggesting potential approaches to overcome the challenges. Data format and usage notes The datasets for the developed cases are available online at https://doi.org/10.5281/zenodo.15720996 . The DICOM files include the treatment plan for each case, TPS, and the corresponding reference MC dose data. The package also contains a TPS‐ and case‐specific user guide for commissioning the MBDCAs, as well as files necessary to replicate the MC simulations. Potential applications The provided datasets and proposed methodology can serve as a commissioning framework for TPSs that employ MBDCAs, as well as a benchmark for brachytherapy researchers using MC methods and MBDCA developers. They also facilitate intercomparisons of MBDCA performance and provide a quality assurance resource for evaluating future TPS software updates.

A simulated annealing‐based Bayesian network structure optimization framework for late morbidity prediction with a large prospective dataset

Abstract Background Bayesian networks are seeing increased usage in healthcare, particularly for modeling complex treatment decisions under uncertainty. Bayesian networks offer significant advantages over classical machine learning and deep learning techniques due to their interpretability, with the network visualized through a directed acyclic graph outlining conditional relationships. Prior clinical knowledge can also be incorporated into these networks to enhance their clarity and facilitate integration into clinical workflows. However, out‐of‐box optimization techniques may produce networks that are not logically coherent or reflective of clinical understanding and may focus solely on optimizing information‐based metrics without consideration for performance metrics crucial for developing predictive models. In late morbidity modeling, where the risk factors surrounding an outcome may be complex, intercorrelated, and not yet fully identified, it is important to have a customizable optimization approach to automatically produce logical, interpretable Bayesian networks that outline these complex outcomes. Purpose Develop a simulated annealing‐based framework for developing Bayesian network structures for late morbidity prediction in cervical cancer patients, addressing limitations of traditional optimization techniques and prioritizing interpretability. Methods This study utilizes the multi‐center EMBRACE I cervical cancer dataset ( n  = 1153) to develop Bayesian network structures for late moderate‐to‐severe (grade ≥2) cystitis (CTCAEv.3) prediction. The dataset was split into training/validation data (80%) and holdout test data (20%). A process of 10 × 5‐fold cross‐validation was integrated into the optimization framework. A simulated annealing‐based optimization method was developed incorporating information—theoretic measures, predictive performance measures, and complexity measures. The different network structures developed by this framework were compared in terms of complexity, interpretability, and predictive performance to optimization methods available out‐of‐box from the PyAgrum package for Python (Greedy Hill Climbing, Tree‐Augmented Naïve Bayes, and Chow‐Liu Optimization). Bayesian networks were also compared to conventional machine learning classifiers in terms of feature importance and predictive performance. Differences in model predictions arising from structure differences were assessed with Cochran's Q ‐test ( p  < 0.05). Results The simulated annealing framework demonstrated the ability to produce Bayesian network structures with comparable or superior predictive performance compared to out‐of‐box models. A statistically significant performance difference was identified between the simulated annealing and out‐of‐box methods with Cochran's Q ‐test ( p  = 0.03). The simulated annealing approach equalled or outperformed out‐of‐box models on a bootstrapped holdout test set, with a balanced accuracy of 64.1%, an F1 macro score of 55.9%, and an ROC‐AUC of 0.66. Simulated annealing models also featured fewer arcs and nodes, with this simplification resulting in networks that were easier to interpret without compromising on predictive performance, highlighting the effectiveness of simulated annealing in creating highly interpretable models for clinical use. Conclusion The proposed simulated annealing‐based framework represents a novel method for automatically generating Bayesian network structures for cervical cancer late morbidity modeling. Compared to out‐of‐box optimization techniques, the simulated annealing Bayesian networks provide comparable or superior predictive performance while constructing a more simple, interpretable network useful for clinical implementation.

CT‐based radiomics prediction of CXCL13 expression in ovarian cancer

AbstractBackgroundOvarian cancer, the most common malignancy in the female reproductive system, and patients tend to be at middle and advanced clinical stages when diagnosed. Therefore, early detection and early diagnosis have important clinical significance for the treatment of ovarian cancer patients. CXCL13, a chemokine with the ligands CXCR3 and CXCR5, is involved in the tumor metastasis process.PurposeThis study aimed to predict mRNA expression of CXCL13 in ovarian cancer tissues noninvasively.MethodsMedical imaging data and transcriptomic sequencing data of the 343 ovarian cancer patients were downloaded from the TCIA and TCGA databases, respectively. Seventy‐six radiomics features were extracted from the CT data. Seven features were selected for model construction by using logistic regression. Accuracy, specificity, sensitivity, positive predictive value, and negative predictive value were used to evaluate the radiomics model.ResultsHigh CXCL13 expression was found to be a significant protective factor for OS [HR (95% CI) = 0.755 (0.622–0.916),p = 0.004]. There was a significant positive correlation between CXCL13 and the degree of eosinophil infiltration. A calibration curve and the Hosmer‒Lemeshow goodness‐of‐fit test showed that the prediction probability of the radiomics prediction model for high expression of CXCL13 was consistent with the true value. The AUC value of the nomogram model's ability to predict OS (12 months) was 0.758. The calibration plot and DCA both showed high clinical applicability for the nomogram model.ConclusionCXCL13 is a candidate predictive biomarker for OC and correlates with the degree of plasma cell and eosinophil infiltration.

POD–Kalman filtering for improving noninvasive 3D temperature monitoring in MR‐guided hyperthermia

AbstractBackgroundDuring resonance frequency (RF) hyperthermia treatment, the temperature of the tumor tissue is elevated to the range of 39–44°C. Accurate temperature monitoring is essential to guide treatments and ensure precise heat delivery and treatment quality. Magnetic resonance (MR) thermometry is currently the only clinical method to measure temperature noninvasively in a volume during treatment. However, several studies have shown that this approach is not always sufficiently accurate for thermal dosimetry in areas with motion, such as the pelvic region. Model‐based temperature estimation is a promising approach to correct and supplement 3D online temperature estimation in regions where MR thermometry is unreliable or cannot be measured. However, complete 3D temperature modeling of the pelvic region is too complex for online usage.PurposeThis study aimed to evaluate the use of proper orthogonal decomposition (POD) model reduction combined with Kalman filtering to improve temperature estimation using MR thermometry. Furthermore, we assessed the benefit of this method using data from hyperthermia treatment where there were limited and unreliable MR thermometry measurements.MethodsThe performance of POD–Kalman filtering was evaluated in several heating experiments and for data from patients treated for locally advanced cervical cancer. For each method, we evaluated the mean absolute error (MAE) concerning the temperature measurements acquired by the thermal probes, and we assessed the reproducibility and consistency using the standard deviation of error (SDE). Furthermore, three patient groups were defined according to susceptibility artifacts caused by the level of intestinal gas motion to assess if the POD–Kalman filtering could compensate for missing and unreliable MR thermometry measurements.ResultsFirst, we showed that this method is beneficial and reproducible in phantom experiments. Second, we demonstrated that the combined method improved the match between temperature prediction and temperature acquired by intraluminal thermometry for patients treated for locally advanced cervical cancer. Considering all patients, the POD–Kalman filter improved MAE by 43% (filtered MR thermometry = 1.29°C, POD–Kalman filtered temperature = 0.74°C). Moreover, the SDE was improved by 47% (filtered MR thermometry = 1.16°C, POD–Kalman filtered temperature = 0.61°C). Specifically, the POD–Kalman filter reduced the MAE by approximately 60% in patients whose MR thermometry was unreliable because of the great amount of susceptibilities caused by the high level of intestinal gas motion.ConclusionsWe showed that the POD–Kalman filter significantly improved the accuracy of temperature monitoring compared to MR thermometry in heating experiments and hyperthermia treatments. The results demonstrated that POD–Kalman filtering can improve thermal dosimetry during RF hyperthermia treatment, especially when MR thermometry is inaccurate.

The population percentile allowance method for determining systematic spatial error tolerances for temporary intensity modulated brachytherapy

AbstractBackgroundMultiple approaches are under development for delivering temporary intensity modulated brachytherapy (IMBT) using partially shielded applicators wherein the delivered dose distributions are sensitive to spatial uncertainties in both the applicator position and shield orientation, rather than only applicator position as with conventional high‐dose‐rate brachytherapy (HDR‐BT). Sensitivity analyses to spatial uncertainties have been reported as components of publications on these emerging technologies, however, a generalized framework for the rigorous determination of the spatial uncertainty tolerances of dose‐volume parameters is needed.PurposeTo derive and present the population percentile allowance (PPA) method, a generalized mathematical and statistical framework to evaluate the tolerance of temporary IMBT approaches to spatial uncertainties in applicator position and shield orientation.MethodsA mathematical formalism describing geometric applicator position and shield orientation shifts was derived that supports straight and curved applicators and applies to serial and helical rotating shield brachytherapy (RSBT) and direction modulated brachytherapy (DMBT). The PPA method entails defining the percentage of a patient population receiving a given therapy that is, allowed to receive dose‐volume errors in the target volume and specified organs at risk of a defined percentage or less, then determining what combinations of applicator position and shield orientation systematic errors would be expected to produce that outcome in the population. The PPA method was applied to the use case of multi‐shield helical 169Yb‐based RSBT for cervical cancer, with 45° and 180° shield emission angles. A total of 37 cervical cancer patients were considered in the population, with average (± 1 standard deviation) HR‐CTV volumes of 79 cm3 ± 37 cm3 and optimized baseline treatment plans (no spatial uncertainties applied) created for each patient to meet dose‐volume requirements of 85 GyEQD2 (equivalent uniform dose in 2 Gy fraction), with D2cc tolerance doses of 90 GyEQD2, 75 GyEQD2, and 75 GyEQD2 for bladder, rectum, and sigmoid colon, respectively.ResultsFor the PPA requirement that 90% of cervical cancer patients receiving multi‐shield helical RSBT could have a maximum dose‐volume uncertainty of 10% for high‐risk clinical target volume (HR‐CTV) D90 (minimum dose to hottest 90%) and bladder, rectum, and sigmoid colon D2cc (minimum dose to hottest 2 cm3), the tolerance systematic applicator position and shield orientation uncertainties were approximately ± 1.0 mm and ± 4.25°, respectively. For ± 1.5 mm and ± 5° systematic applicator position and shield orientation tolerances, 90% of the patients considered would have a maximum dose‐volume uncertainty of 12.8% or less.ConclusionThe PPA method was formalized to determine the temporary IMBT spatial uncertainty tolerances that would be expected to result in an allowed percentage of a population of patients receiving relative dose‐volume errors above a defined percentage. Multi‐shield, helical 169Yb‐based RSBT for cervical cancer was evaluated and tolerances determined, which, if applied on each treatment fraction, would represent an extreme situation. The PPA method is applicable to a variety of temporary IMBT approaches and can be used to rigorously determine the design parameters for the delivery systems such as mechanical driver motor accuracy, shield angle backlash, applicator rotation, and applicator fixation stability.

Delineation of clinical target volume and organs at risk in cervical cancer radiotherapy by deep learning networks

AbstractPurposeDelineation of the clinical target volume (CTV) and organs‐at‐risk (OARs) is important in cervical cancer radiotherapy. But it is generally labor‐intensive, time‐consuming, and subjective. This paper proposes a parallel‐path attention fusion network (PPAF‐net) to overcome these disadvantages in the delineation task.MethodsThe PPAF‐net utilizes both the texture and structure information of CTV and OARs by employing a U‐Net network to capture the high‐level texture information, and an up‐sampling and down‐sampling (USDS) network to capture the low‐level structure information to accentuate the boundaries of CTV and OARs. Multi‐level features extracted from both networks are then fused together through an attention module to generate the delineation result.ResultsThe dataset contains 276 computed tomography (CT) scans of patients with cervical cancer of staging IB‐IIA. The images are provided by the West China Hospital of Sichuan University. Simulation results demonstrate that PPAF‐net performs favorably on the delineation of the CTV and OARs (e.g., rectum, bladder and etc.) and achieves the state‐of‐the‐art delineation accuracy, respectively, for the CTV and OARs. In terms of the Dice Similarity Coefficient (DSC) and the Hausdorff Distance (HD), 88.61% and 2.25 cm for the CTV, 92.27% and 0.73 cm for the rectum, 96.74% and 0.68 cm for the bladder, 96.38% and 0.65 cm for the left kidney, 96.79% and 0.63 cm for the right kidney, 93.42% and 0.52 cm for the left femoral head, 93.69% and 0.51 cm for the right femoral head, 87.53% and 1.07 cm for the small intestine, and 91.50% and 0.84 cm for the spinal cord.ConclusionsThe proposed automatic delineation network PPAF‐net performs well on CTV and OARs segmentation tasks, which has great potential for reducing the burden of radiation oncologists and increasing the accuracy of delineation. In future, radiation oncologists from the West China Hospital of Sichuan University will further evaluate the results of network delineation, making this method helpful in clinical practice.

Attention feature fusion methodology with additional constraint for ovarian lesion diagnosis on magnetic resonance images

AbstractPurposeIt is challenging for radiologists and gynecologists to identify the type of ovarian lesions by reading magnetic resonance (MR) images. Recently developed convolutional neural networks (CNNs) have made great progress in computer vision, but their architectures still need modification if they are used in processing medical images. This study aims to improve the feature extraction capability of CNNs, thus promoting the diagnostic performance in discriminating between benign and malignant ovarian lesions.MethodsWe introduce a feature fusion architecture and insert the attention models in the neural network. The features extracted from different middle layers are integrated with reoptimized spatial and channel weights. We add a loss function to constrain the additional probability vector generated from the integrated features, thus guiding the middle layers to emphasize useful information. We analyzed 159 lesions imaged by dynamic contrast‐enhanced MR imaging (DCE‐MRI), including 73 benign lesions and 86 malignant lesions. Senior radiologists selected and labeled the tumor regions based on the pathology reports. Then, the tumor regions were cropped into 7494 nonoverlapping image patches for training and testing. The type of a single tumor was determined by the average probability scores of the image patches belonging to it.ResultsWe implemented fivefold cross‐validation to characterize our proposed method, and the distribution of performance matrics was reported. For all the test image patches, the average accuracy of our method is 70.5% with an average area under the curve (AUC) of 0.785, while the baseline is 69.4% and 0.773, and for the diagnosis of single tumors, our model achieved an average accuracy of 82.4% and average AUC of 0.916, which were better than the baseline (81.8% and 0.899). Moreover, we evaluated the performance of our proposed method utilizing different CNN backbones and different attention mechanisms.ConclusionsThe texture features extracted from different middle layers are crucial for ovarian lesion diagnosis. Our proposed method can enhance the feature extraction capabilities of different layers of the network, thereby improving diagnostic performance.

Source position measurement by Cherenkov emission imaging from applicators for high‐dose‐rate brachytherapy

PurposeWe developed a novel and simple method to measure the source positions in applicators directly for high‐dose‐rate (HDR) brachytherapy based on Cherenkov emission imaging, and evaluated the performance.MethodsThe light emission from plastic applicators used in cervical cancer treatments, irradiated by an 192Ir γ‐ray source, was captured using a charge‐coupled device camera. Moreover, we attached plastics of different shapes, including tapes, tubes, and plates to a metal applicator, to use as screens for the Cherenkov imaging. We determined the source positions and dwell intervals from the light profiles along with the applicator and compared these with preset values and dummy marker measurements.ResultsThe source positions and dwell intervals measured from the light images were comparable to the dummy marker measurements and preset values. The distance from the applicator tip to the first source positions agreed with the dummy marker measurements within 0.2 mm for the plastic tandem. The dwell intervals measured using the Cherenkov method agreed with the preset values within 0.6 mm. The distances measured with three plastic types on the metal applicator also agreed with the dummy marker measurements within 0.2 mm. The dwell intervals measured using the plastic tape agreed with the preset values within 0.7 mm.ConclusionsThe proposed method should be suitable for rapid and easy quality assurance (QA) investigations in HDR brachytherapy, as it enables source position using a single image. The method allows for real‐time, filmless measurements of the source positions to be obtained and is useful for rapid feedback in QA procedures.

Automatic contouring system for cervical cancer using convolutional neural networks

PurposeTo develop a tool for the automatic contouring of clinical treatment volumes (CTVs) and normal tissues for radiotherapy treatment planning in cervical cancer patients.MethodsAn auto‐contouring tool based on convolutional neural networks (CNN) was developed to delineate three cervical CTVs and 11 normal structures (seven OARs, four bony structures) in cervical cancer treatment for use with the Radiation Planning Assistant, a web‐based automatic plan generation system. A total of 2254 retrospective clinical computed tomography (CT) scans from a single cancer center and 210 CT scans from a segmentation challenge were used to train and validate the CNN‐based auto‐contouring tool. The accuracy of the tool was evaluated by calculating the Sørensen‐dice similarity coefficient (DSC) and mean surface and Hausdorff distances between the automatically generated contours and physician‐drawn contours on 140 internal CT scans. A radiation oncologist scored the automatically generated contours on 30 external CT scans from three South African hospitals.ResultsThe average DSC, mean surface distance, and Hausdorff distance of our CNN‐based tool were 0.86/0.19 cm/2.02 cm for the primary CTV, 0.81/0.21 cm/2.09 cm for the nodal CTV, 0.76/0.27 cm/2.00 cm for the PAN CTV, 0.89/0.11 cm/1.07 cm for the bladder, 0.81/0.18 cm/1.66 cm for the rectum, 0.90/0.06 cm/0.65 cm for the spinal cord, 0.94/0.06 cm/0.60 cm for the left femur, 0.93/0.07 cm/0.66 cm for the right femur, 0.94/0.08 cm/0.76 cm for the left kidney, 0.95/0.07 cm/0.84 cm for the right kidney, 0.93/0.05 cm/1.06 cm for the pelvic bone, 0.91/0.07 cm/1.25 cm for the sacrum, 0.91/0.07 cm/0.53 cm for the L4 vertebral body, and 0.90/0.08 cm/0.68 cm for the L5 vertebral bodies. On average, 80% of the CTVs, 97% of the organ at risk, and 98% of the bony structure contours in the external test dataset were clinically acceptable based on physician review.ConclusionsOur CNN‐based auto‐contouring tool performed well on both internal and external datasets and had a high rate of clinical acceptability.

A novel minimally invasive dynamic‐shield, intensity‐modulated brachytherapy system for the treatment of cervical cancer

PurposeTo present a novel, MRI‐compatible dynamicshield intensity modulated brachytherapy (IMBT) applicator and delivery system using 192Ir, 75Se, and 169Yb radioisotopes for the treatment of locally advanced cervical cancer. Needle‐free IMBT is a promising technique for improving target coverage and organs at risk (OAR) sparing.Methods and materialsThe IMBT delivery system dynamically controls the rotation of a novel tungsten shield placed inside an MRI‐compatible, 6‐mm wide intrauterine tandem. Using 36 cervical cancer cases, conventional intracavitary brachytherapy (IC‐BT) and intracavitary/interstitial brachytherapy (IC/IS‐BT) (10Ci 192Ir) plans were compared to IMBT (10Ci 192Ir; 11.5Ci 75Se; 44Ci 169Yb). All plans were generated using the Geant4‐based Monte Carlo dose calculation engine, RapidBrachyMC. Treatment plans were optimized then normalized to the same high‐risk clinical target volume (HR‐CTV) D90 and the D2cc for bladder, rectum, and sigmoid in the research brachytherapy planning system, RapidBrachyMCTPS. Plans were renormalized until either of the three OAR reached dose limits to calculate the maximum achievable HR‐CTV D90 and D98.ResultsCompared to IC‐BT, IMBT with either of the three radionuclides significantly improves the HR‐CTV D90 and D98 by up to 5.2% ± 0.3% (P < 0.001) and 6.7% ± 0.5% (P < 0.001), respectively, with the largest dosimetric enhancement when using 169Yb followed by 75Se and then 192Ir. Similarly, D2cc for all OAR improved with IMBT by up to 7.7% ± 0.6% (P < 0.001). For IC/IS‐BT cases, needle‐free IMBT achieved clinically acceptable plans with 169Yb‐based IMBT further improving HR‐CTV D98 by 1.5% ± 0.2% (P = 0.034) and decreasing sigmoid D2cc by 1.9% ± 0.4% (P = 0.048). Delivery times for IMBT are increased by a factor of 1.7, 3.3, and 2.3 for 192Ir, 75Se, and 169Yb, respectively, relative to conventional 192Ir BT.ConclusionsDynamic shield IMBT provides a promising alternative to conventional IC‐ and IC/IS‐BT techniques with significant dosimetric enhancements and even greater improvements with intermediate energy radionuclides. The ability to deliver a highly conformal, OAR‐sparing dose without IS needles provides a simplified method for improving the therapeutic ratio less invasively and in a less resource intensive manner.

A biomechanical finite element model to generate a library of cervix CTVs

PurposeTo generate a series of physiologically plausible cervix CTVs by biomechanically modeling organ deformation as a consequence of bladder filling. This series can serve as planning CTVs for radiotherapy treatment of cervical cancer patients using a library of plans (LoP) strategy.MethodsThe model was constructed based on the full and empty bladder scans of 20 cervical cancer patients, for which the bladder, rectum and the clinical target volume (CTV) of the cervix were delineated. Finite element modeling (FEM) was used to deform empty to full bladder anatomy. This deformation comprised two steps. In the first step, the surfaces of the bladder and rectum of the empty bladder anatomy were explicitly deformed to the full bladder anatomy and imported as enforced displacements into the biomechanical model. These surface displacements cause volumetric deformations of the bladder, rectum and cervix CTV meshes, dictated by their respective elastic properties and the type of contact among them. In the second step, the residual offset between the simulated and target CTV was corrected by an additional thin plate spline warp. Intermediate structural outputs of a linear superposition of the biomechanical and residual warp then constituted the library of CTVs for each patient. The residual warp was minimized by optimizing the FEM parameters over the 20 patients. Finally, the model was tested for nine healthy volunteers for which repeat MR scans were available as the bladder filled from empty to full. Small and large movers were identified depending on the extent of CTV motion, and analyzed separately. The proposed method was compared against the method currently used in our institute, in which intermediate structures are linearly interpolated between full and empty bladder anatomy, using a thin plate spline warp. The comparison metrics used were the ability to preserve CTV volume throughout the deformation, and residual offsets between repeat and library CTV.ResultsOptimal model parameters were found to be compatible with published values. While for the current method, the median CTV volume shrunk by 4% for large movers halfway the deformation (and by up to 10% for individual cases), the proposed FEM‐based method preserved CTV volumes throughout the deformation. Regional residual errors between repeat and library CTV reduced by up to 3 mm when averaged over the group of large movers. For individual cases this regional error reduction could be as large as 8 mm.ConclusionsWe developed a robust and automatic method to create a patient‐specific FEM‐based LoP. The FEM‐based method resulted in more accurate library of planning CTVs as compared to the current method, with the greatest improvements observed for patients with large CTV motion. The biomechanical model simulates volumetric deformations from empty to full bladder anatomy, paving the way for dose accumulation in an LoP setting.

Needle‐free cervical cancer treatment using helical multishield intracavitary rotating shield brachytherapy with the 169Yb Isotope

PurposeTo assess the capability of an intracavitary 169Yb‐based helical multishield rotating shield brachytherapy (RSBT) delivery system to treat cervical cancer. The proposed RSBT delivery system contains a pair of 1.25 mm thick platinum partial shields with 45° and 180° emission angles, which travel in a helical pattern within the applicator.MethodsA helically threaded tandem applicator with a 45° tandem curvature containing a helically threaded catheter was designed. A 0.6 mm diameter 169Yb source with a length of 10.5 mm was simulated. A 37‐patient treatment planning study, based on Monte Carlo dose calculations using MCNP5, was conducted with high‐risk clinical target volumes (HR‐CTVs) of 41.2–192.8 cm3 (average ± standard deviation of 79.9 ± 35.8 cm3). All patients were assumed to receive 25 fractions of 1.8 Gy of external beam radiation therapy (EBRT) before receiving 5 fractions of high‐dose‐rate brachytherapy (HDR‐BT). For each patient, 192Ir‐based intracavitary (IC) HDR‐BT, 192Ir‐based intracavitary/interstitial (IC/IS) HDR‐BT using a hybrid applicator with eight IS needles, and 169Yb‐based RSBT plans were generated.ResultsFor the IC, IC/IS, and RSBT treatment plans, 38%, 84%, and 86% of the plans, respectively, met the planning goal of an HR‐CTV D90 (minimum dose to hottest 90%) of 85 GyEQD2 (α/β = 10 Gy). Median (25th percentile, 75th percentile) treatment times for IC, IC/IS, and RSBT were 11.71 (6.62, 15.40) min, 68.00 (45.02, 80.02) min, and 25.30 (13.87, 35.39) min, respectively. 192Ir activities ranging from 159.1–370 GBq (4.3–10 Ci) and 169Yb activities ranging from 429.2–999 GBq (11.6–27 Ci) were used, which correspond to the same clinical ranges of dose rates at 1 cm off‐source‐axis in water. Extra needle insertion and planning time beyond that needed for intracavitary‐only approaches was accounted for in the IC/IS treatment time calculations.Conclusion169Yb‐based RSBT for cervical cancer met the HR‐CTV D90 goal of 85 Gy in a greater percentage of the patients considered than IC/IS (86% vs 84%, respectively) and can reduce overall treatment time relative to IC/IS. 169Yb‐based RSBT could be used to replace IC/IS in instances where IC/IS treatment is not available, especially in instances when HR‐CTV volumes are ≥30 cm3.

MRI‐based radiomics nomogram for the preoperative prediction of deep myometrial invasion of FIGO stage I endometrial carcinoma

AbstractBackgroundEndometrial carcinoma (EC) is one of the most common gynecological malignancies with an increasing incidence, and an accurate preoperative diagnosis of deep myometrial invasion (DMI) is crucial for personalized treatment.ObjectiveTo determine the predictive value of a magnetic resonance imaging (MRI)‐based radiomics nomogram for the presence of DMI in the International Federation of Gynecology and Obstetrics (FIGO) stage I EC.MethodsWe retrospectively collected 163 patients with pathologically confirmed stage I EC from two centers and divided all samples into a training group (Center 1) and a validation group (Center 2). Clinical and routine imaging indicators were analyzed by logistical regression to construct a conventional diagnostic model (M1). Radiomics features extracted from the axial T2‐weighted and axial contrast‐enhanced T1‐weighted (CE‐T1W) images were treated with the intraclass correlation coefficient, Mann–Whitney U test, least absolute shrinkage and selection operator, and logistic regression analysis with Akaike information criterion to build a combined radiomics signature (M2). A nomogram (M3) was constructed by M1 and M2. Calibration and decision curves were drawn to evaluate the nomogram in the training and validation cohorts. The diagnostic performance of each indicator and model was evaluated by the area under the receiver operating characteristic curve (AUC).ResultThe four most significant radiomics features were finally selected from the CE‐T1W MRI. For the diagnosis of DMI, the AUCT/AUCV of M1 was 0.798/0.738, the AUCT/AUCV of M2 was 0.880/0.852, and the AUCT/AUCV of M3 was 0.936/0.871 in the training and validation groups, respectively. The calibration curves showed that M3 was in good agreement with the ideal values. The decision curve analysis suggested potential clinical application values of the nomogram.ConclusionA nomogram based on MRI radiomics and clinical imaging indicators can improve the diagnosis of DMI in patients with FIGO stage I EC.