Journal of Applied Clinical Medical Physics

A deep learning‐based 3D Prompt‐nnUnet model for automatic segmentation in brachytherapy of postoperative endometrial carcinoma

AbstractPurposeTo create and evaluate a three‐dimensional (3D) Prompt‐nnUnet module that utilizes the prompts‐based model combined with 3D nnUnet for producing the rapid and consistent autosegmentation of high‐risk clinical target volume (HR CTV) and organ at risk (OAR) in high‐dose‐rate brachytherapy (HDR BT) for patients with postoperative endometrial carcinoma (EC).Methods and materialsOn two experimental batches, a total of 321 computed tomography (CT) scans were obtained for HR CTV segmentation from 321 patients with EC, and 125 CT scans for OARs segmentation from 125 patients. The numbers of training/validation/test were 257/32/32 and 87/13/25 for HR CTV and OARs respectively. A novel comparison of the deep learning neural network 3D Prompt‐nnUnet and 3D nnUnet was applied for HR CTV and OARs segmentation. Three‐fold cross validation and several quantitative metrics were employed, including Dice similarity coefficient (DSC), Hausdorff distance (HD), 95th percentile of Hausdorff distance (HD95%), and intersection over union (IoU).ResultsThe Prompt‐nnUnet included two forms of parameters Predict‐Prompt (PP) and Label‐Prompt (LP), with the LP performing most similarly to the experienced radiation oncologist and outperforming the less experienced ones. During the testing phase, the mean DSC values for the LP were 0.96 ± 0.02, 0.91 ± 0.02, and 0.83 ± 0.07 for HR CTV, rectum and urethra, respectively. The mean HD values (mm) were 2.73 ± 0.95, 8.18 ± 4.84, and 2.11 ± 0.50, respectively. The mean HD95% values (mm) were 1.66 ± 1.11, 3.07 ± 0.94, and 1.35 ± 0.55, respectively. The mean IoUs were 0.92 ± 0.04, 0.84 ± 0.03, and 0.71 ± 0.09, respectively. A delineation time < 2.35 s per structure in the new model was observed, which was available to save clinician time.ConclusionThe Prompt‐nnUnet architecture, particularly the LP, was highly consistent with ground truth (GT) in HR CTV or OAR autosegmentation, reducing interobserver variability and shortening treatment time.

Clinical target volume (CTV) automatic delineation using deep learning network for cervical cancer radiotherapy: A study with external validation

AbstractPurposeTo explore the accuracy and feasibility of a proposed deep learning (DL) algorithm for clinical target volume (CTV) delineation in cervical cancer radiotherapy and evaluate whether it can perform well in external cervical cancer and endometrial cancer cases for generalization validation.MethodsA total of 332 patients were enrolled in this study. A state‐of‐the‐art network called ResCANet, which added the cascade multi‐scale convolution in the skip connections to eliminate semantic differences between different feature layers based on ResNet‐UNet. The atrous spatial pyramid pooling in the deepest feature layer combined the semantic information of different receptive fields without losing information. A total of 236 cervical cancer cases were randomly grouped into 5‐fold cross‐training (n = 189) and validation (n = 47) cohorts. External validations were performed in a separate cohort of 54 cervical cancer and 42 endometrial cancer cases. The performances of the proposed network were evaluated by dice similarity coefficient (DSC), sensitivity (SEN), positive predictive value (PPV), 95% Hausdorff distance (95HD), and oncologist clinical score when comparing them with manual delineation in validation cohorts.ResultsIn internal validation cohorts, the mean DSC, SEN, PPV, 95HD for ResCANet achieved 74.8%, 81.5%, 73.5%, and 10.5 mm. In external independent validation cohorts, ResCANet achieved 73.4%, 72.9%, 75.3%, 12.5 mm for cervical cancer cases and 77.1%, 81.1%, 75.5%, 10.3 mm for endometrial cancer cases, respectively. The clinical assessment score showed that minor and no revisions (delineation time was shortened to within 30 min) accounted for about 85% of all cases in DL‐aided automatic delineation.ConclusionsWe demonstrated the problem of model generalizability for DL‐based automatic delineation. The proposed network can improve the performance of automatic delineation for cervical cancer and shorten manual delineation time at no expense to quality. The network showed excellent clinical viability, which can also be even generalized for endometrial cancer with excellent performance.

A novel multi‐channel applicator with a U‐shaped channel for vaginal intracavity brachytherapy

AbstractBackgroundEndometrial cancer is one of the most common gynecological malignancies in the world. Vaginal brachytherapy is an important postoperative adjuvant treatment for endometrial cancer. However, a common problem with existing applicators is insufficient dose at the vaginal apex.PurposeThis study describes the Hangzhou (HZ) cylinder, a novel 3D printed vaginal intracavity brachytherapy applicator, detailing its characteristics, dose distribution, and clinical applications.Methods and MaterialsThe HZ cylinder is distinguished by its unique structure: a U‐shaped channel with a 2 mm diameter, a straight central axis channel of the same diameter, and 10 parallel straight channels. For comparison, standard plans were employed, designed to ensure that a minimum of 95% of the prescribed dose reached 5 mm beneath the mucosal surface. We conducted comparative analyses of mucosal surface doses and doses at a 5 mm depth below the mucosa between the HZ cylinder and a conventional single‐channel cylinder across various treatment schemes. Additionally, the study examined dose differences in target volume and organs at risk (OARs) between actual HZ cylinder plans and hypothetical single‐channel plans.ResultsIn the standard plans, mucosal surface doses at the apex of the vagina were 209.32% and 200.61% of the prescribed dose with the HZ and single‐channel cylinders, respectively. The doses on the left and right wall mucosal surfaces varied from 149.26% to 178.13% and 142.98% to 180.75% of the prescribed dose, and on the anterior and posterior wall mucosal surfaces varied from 128.87% to 138.50% and 142.98% to 180.75% of the prescribed dose. Analysis of 24 actual treatment plans revealed that when the vaginal tissue volume dose covering 98% (vaginal D98%) was comparable between the HZ cylinder and virtual single‐channel plans (6.74 ± 0.07 Gy vs. 6.69 ± 0.10 Gy, p = 0.24), rectum doses of HZ cylinder plans were significantly lower than those of single‐channel plans (D1cc, 5.96 ± 0.56 Gy vs. 6.26 ± 0.71 Gy, p = 0.02 and D2cc, 5.26 ± 0.52 Gy vs. 5.56 ± 0.62 Gy, p = 0.02).ConclusionsThe HZ cylinder demonstrates a reduction in dose to the rectum and bladder while maintaining adequate target volume coverage. Its mucosal surface dose is comparable to that of the traditional single‐channel cylinder. These findings suggest that the HZ cylinder is a viable and potentially safer alternative for vaginal brachytherapy, warranting further investigation with larger sample sizes.

The feasibility study on the generalization of deep learning dose prediction model for volumetric modulated arc therapy of cervical cancer

AbstractPurposeTo develop a 3D‐Unet dose prediction model to predict the three‐dimensional dose distribution of volumetric modulated arc therapy (VMAT) for cervical cancer and test the dose prediction performance of the model in endometrial cancer to explore the feasibility of model generalization.MethodsOne hundred and seventeen cases of cervical cancer and 20 cases of endometrial cancer treated with VMAT were used for the model training, validation, and test. The prescribed dose was 50.4 Gy in 28 fractions. Eight independent channels of contoured structures were input to the model, and the dose distribution was used as the output of the model. The 3D‐Unet prediction model was trained and validated on the training set (n = 86) and validation set (n = 11), respectively. Then the model was tested on the test set (n = 20) of cervical cancer and endometrial cancer, respectively. The results between clinical dose distribution and predicted dose distribution were compared in the following aspects: (a) the mean absolute error (MAE) within the body, (b) the Dice similarity coefficients (DSCs) under different isodose volumes, (c) the dosimetric indexes including the mean dose (Dmean), the received dose of 2 cm3 (D2cc), the percentage volume of receiving 40 Gy dose of organs‐at‐risk (V40), planning target volume (PTV) D98%, and homogeneity index (HI), (d) dose–volume histograms (DVHs).ResultsThe model can accurately predict the dose distribution of the VMAT plan for cervical cancer and endometrial cancer. The overall average MAE and maximum MAE for cervical cancer were 2.43 ± 3.17% and 3.16 ± 4.01% of the prescribed dose, respectively, and for endometrial cancer were 2.70 ± 3.54% and 3.85 ± 3.11%. The average DSCs under different isodose volumes is above 0.9. The predicted dosimetric indexes and DVHs are equivalent to the clinical dose for both cervical cancer and endometrial cancer, and there is no statistically significant difference.ConclusionA 3D‐Unet dose prediction model was developed for VMAT of cervical cancer, which can predict the dose distribution accurately for cervical cancer. The model can also be generalized for endometrial cancer with good performance.

Analysis of the effect of dual reference lines on first positioning accuracy in intensity‐modulated radiotherapy for cervical cancers

Abstract Objective To investigate the use of dual‐reference lines to reduce positioning errors and optimize modified planning target volume (MPTV) in volume modulated arc therapy (VMAT) for cervical cancers. Methods Thirty‐seven patients with FIGO stage IIB‐IVA cervical cancer and no distant metastasis, who underwent radiotherapy in a tertiary hospital from June 15 2022 to September 15 2023, were selected and randomly divided into dual positioning reference line group (dual‐line group, 21 cases) and single positioning reference line group (single‐line group, 16 cases). A single reference line was made in pelvic region in single‐line group, while dual reference lines were made in stable abdominopelvic region in dual‐line group. The cone‐beam computed tomography (CBCT) was conducted to determine the positioning errors and calculate MPTV. Results Linear error in Y direction, rotational errors in the rotation around Y ‐axis ( RY )/ Z ‐axis ( RZ ) (0.27 ± 0.12 cm, 0.60 ± 0.42°, 0.48 ± 0.44°) in dual‐line group were smaller than those (0.35 ± 0.22 cm, 0.78 ± 0.45°, 0.85 ± 0.66°) in single‐line group ( p  < 0.05). The thresholds of 0.4 cm and 1.4° were set as the boundary values for linear and rotational errors, respectively. There were significant statistical differences in the distribution of positioning errors in the six directions between the two groups ( p  < 0.001), with higher positioning error rates in the Y (77.78%), RY (41.46%), and RZ (53.66%) directions, respectively. Median total positioning time in dual‐line group (8.27 min, interquartile rang [IQR]: 7.65–8.63) was shorter than that (8.75 min, IQR: 7.89–9.45) in single‐line group (Z = 3.53998, p  < 0.001). MPTVs in X , Y , and Z directions (0.25, 0.37, and 0.10 cm) in dual‐line group were smaller than those (0.31, 0.56, and 0.11 cm) in single‐line group. Conclusion Dual‐reference lines improve positioning accuracy, reduce MPTV, and enhance efficiency in VMAT for middle and advanced cervical cancers, offering a clinically practical solution for precision radiotherapy.

Simultaneous integrated boost in locally advanced cervical cancer patients ineligible for brachytherapy: Dosimetric comparison of VMAT versus helical tomotherapy

Abstract Objective This retrospective, exploratory study evaluated the dosimetric comparison and feasibility of simultaneous integrated boost (SIB) delivered with volumetric modulated arc therapy (VMAT) and helical tomotherapy (HT) in patients with locally advanced cervical cancer ineligible for brachytherapy (BRT). Methods This exploratory study involved dosimetric treatment planning based on data from 10 patients diagnosed with locally advanced cervical cancer. SIB plans delivering 77.5 Gy to the primary tumor and 45 Gy to elective regions in 25 fractions were generated using VMAT and HT techniques. SIB plans were created using the VMAT technique in the Monaco 5.51 treatment planning system (TPS) (Elekta, Stockholm, Sweden) and the HT technique in the Precision 3.3.1 TPS (Accuray, Madison, Wisconsin, USA). Planning target volume (PTV) coverage, organ‐at‐risk (OAR) doses, and dose–volume metrics of SIB plans were compared between the VMAT and HT techniques using appropriate statistical tests. Results The median PTV77.5 volume was 83.4 cc (range: 14.9–218.2 cc). For PTV77.5, HT yielded a higher mean dose (79.52 Gy vs. 77.54 Gy,  p  = 0.001) and a superior conformity index (0.81 vs. 0.71,  p  < 0.001). Bladder dose metrics were significantly lower with HT, including V53.28 Gy (5.66% vs. 10.31%; p  = 0.006) and D 2cc (66.47 Gy vs. 70.60 Gy;  p  < 0.05), while bowel V36.89 Gy volume was also reduced (94.03 cc vs. 115.13 cc;  p  = 0.038). No statistically significant differences were observed in rectal, sigmoid, or femoral head dose parameters. Beam‐on time was significantly longer for HT than for VMAT (11.0 min vs. 3.4 min,  p  = 0.005). Conclusions BRT remains the standard of care for cervical cancer. For patients ineligible for BRT, SIB is dosimetrically feasible; in this exploratory planning study, HT‐SIB provided better target coverage and OAR sparing than VMAT‐SIB. Prospective multicenter validation is necessary before wider clinical adoption.

Analysis of treatment planning time and optimization parameters for inverse planning for intracavitary and interstitial brachytherapy in uterine cervical cancer

Abstract Purpose This study aimed to investigate the effect of inverse planning parameters on dose‐volume indices in brachytherapy for uterine cervical cancer. Methods Fourteen consecutive patients with cervical cancer who received intracavitary and interstitial brachytherapy (IC/ISBT) were selected. Tandem, ovoid, and interstitial needles were used in all cases. The evaluation plans were recalculated from the first fraction of clinical brachytherapy plans. The correlation between the 11 dose optimization parameters of inverse planning and the 13 dose‐volume indices was evaluated. The parameters were adjusted in five levels, and dose optimization was performed in hybrid inverse planning optimization (HIPO). Spearman's rank correlation and multiple regression analyses were conducted to assess the association between the parameters and the indices. The indices included clinical target volume (CTV) dose, organ‐at‐risk (OAR) dose, homogeneity, and conformity. Additionally, the correlation between optimization parameters and calculation time was investigated, along with a technique for efficiently generating treatment plans. Results “CTV Max Weight” and “OAR Max Weight” were the key parameters significantly affecting the indices. Increasing “CTV Max Weight” improved homogeneity but reduced the target coverage. The effect of “OAR Max Weight” on the dose reduction of CTV HR D 90 (β = −0.59) was more significant than that on the dose reduction of OAR D 2cc (β = −0.21). In addition, adjusting “CTV Min Weight” and “CTV Volume” could reduce the hyper‐dose sleeve without increasing the OAR dose. A large number of normal tissue sampling points could negatively affect the dose distributions and increase the calculation times. Conclusion “CTV Max Weight” and “OAR Max Weight” were the most influential parameters in HIPO, significantly affecting dose‐volume indices in IC/ISBT for uterine cervical cancer. Additionally, parameters that regulate the hyper‐dose sleeve and needle‐delivered dose were identified. The quality of treatment planning can be maintained and planning time reduced by appropriately optimizing these parameters.

Automatic radiotherapy planning for deliverable plans using deep learning dose prediction and dose rings optimization in cervical cancer

Abstract Background Automatic radiotherapy (RT) planning based on deep learning (DL) has been extensively researched. However, it is challenging to import the predicted dose distribution into mainstream treatment planning systems (TPSs) and generate clinically deliverable plans. Purpose To investigate the feasibility and accuracy of an automatic volumetric modulated arc therapy (VMAT) and intensity‐modulated radiation therapy (IMRT) planning method for generation of universally deliverable plans based on DL dose prediction and dose rings optimization. Methods First, dose distributions were predicted using a three‐dimensional (3D) Fusion Residual Unet (F‐ResUnet) DL network with data from two hospitals, which included 230 and 210 gynecological cancer (GC) patients underwent VMAT and IMRT, respectively. Then, the predicted dose distributions were discretized into dose rings to optimize the plans automatically in two mainstream TPSs based on the dose rings. Finally, the deliverability of generated plans was verified with patient‐specific quality assurance (PSQA). Results The predicted dose distributions were clinically acceptable with a target coverage over 95%. Compared with the clinical plans, the automatic plans optimized with dose rings achieved a similar dose coverage on planning target volumes (PTV) with an average target coverage over 96.5%. For organs at risk (OARs) sparing, automatic VMAT plans markedly decreased the V 30Gy of left femoral head ( p  = 0.05), right femoral head ( p  = 0.004), and small intestine ( p  = 0.04). The V 45Gy of bladder and rectum in the automatic IMRT plans were reduced by approximately 7% and 9%, respectively. Deliverability verification with PSQA achieved a mean gamma passing rate of 99.1%, 97.1% and 98.3%, 95.0% under the criteria of 3%/3 mm and 3%/2 mm for VMAT and IMRT plans, respectively. Conclusions The proposed automatic planning method combining DL dose prediction and dose rings optimization was feasible to generate universally deliverable VMAT and IMRT plans for gynecological cancer (GC) patients.

Validating clinical feasibility of MRCAT and deep learning‐based synthetic CT images for cervical cancer patient

Abstract Background Various methods have been developed to generate synthetic computed tomography (CT) images from magnetic resonance (MR) images, including segmentation‐based approach with MR calculating attenuation (MRCAT) and deep learning (DL)‐based approach. Purpose In this study, we aimed to validate the conventional radiotherapy (RT) planning process with MRCAT and DL‐based synthetic CT images for five patients with cervical cancer. Methods DL‐based synthetic CT images of the five patients were inferred using a network trained with 40 pairs of CT and deformed, normalized T2‐weighted MR scans; MRCAT images were obtained from mDixon sequences for the tested cases only. On the synthetic CT images, the contouring process for organs‐at‐risk (OARs) was automatically performed with minor adjustments, while two experienced radiation oncologists defined target volumes. Simultaneous integrated boost plans (2.2/2.0/1.8 Gy with 25 fractions) were produced from a commercial treatment planning system (TPS) TomoTherapy. Results The plans with two synthetic CT images were compared with those based on genuine CT images for the five test cases. High geometric similarity was confirmed for the planning target volume (PTV), with average dice similarity coefficient (DSC) of 0.844 for the DL‐based and 0.829 for the MRCAT images. The mean percentage difference in gross tumor volume (GTV) was 20.71 34.28% for DL‐based synthetic CT and 30.31 46.20% for MRCAT images. By contrast, PTV, encompassing GTV, exhibited minimal changes with an average increase of 0.37 3.10% and 1.66 7.62%, respectively. MRCAT images and DL‐based synthetic CT revealed significant differences, relative to true CT images, in the entire volume ( p = 0.03 ) of the bladder and in V 20Gy and V 30Gy of the resultant plans for the bladder ( p = 0.029 and 0.063 ), all plans generated on the synthetic CTs were clinically acceptable and met institutional for target coverage. Conclusion MRCAT and DL‐based synthetic CT images demonstrated clinical applicability, achieving plan quality similar to that of plans based on genuine planning CT images.

Effects of simultaneous multislice acceleration on the stability of radiomics features in parametric maps of IVIM and DKI in uterine cervical cancer

AbstractPurposeThe aim of this study was to investigate the influence of the simultaneous multislice acceleration (SMS) technique as well as two‐dimensional (2D) and three‐dimensional (3D) tumor segmentations on radiomics features (RFs) within the parametric maps of cervical cancer, which were computed by intravoxel incoherent motion (IVIM) and diffusion kurtosis imaging (DKI). Additionally, the study sought to identify those RFs that could characterize the clinical stages (low‐stage vs. high‐stage) of cervical cancer.Materials and methodsMulti‐b‐value diffusion‐weighted imaging (DWI) of 40 patients with cervical cancer were collected using the SMS technique with acceleration factors (AF) of 1–3. RFs were extracted from parametric maps representing pure diffusion coefficient (D), pseudodiffusion coefficient (D*), perfusion fraction (f), mean diffusivity (MD), and mean kurtosis (MK). A total of 93 2D and 93 3D RFs were extracted from per parametric map. The concordance correlation coefficient (CCC) and coefficients of variation (COV) were used to jointly assess the stability of features. Finally, the intra‐class correlation coefficient (ICC) was used for intra‐group consistency assessment. Receiver operating characteristic (ROC) curve was used to evaluate diagnostic performance of stable features in distinguishing lower and higher stages.ResultsFeature stability decreased with higher AF. Among these features, 9.1% of 2D and 12.7% of 3D RFs were stable (CCC > 0.9 and COV ≤ 0.1). ADC maps had the highest stability, whileas D* and f maps had the lowest stability and 3D features were more stable than 2D features. A total of 5 2D and 25 3D stable features were identified that could distinguish lower and higher stages (AUC = 0.66–0.83).ConclusionSMS demonstrated impact on the stability of RFs in IVIM and DKI parametric maps, particularly for D* and f maps. Multi‐b‐value DWI with SMS (AF = 2) was recommended for clinical radiomics research.

Attention 3D UNET for dose distribution prediction of high‐dose‐rate brachytherapy of cervical cancer: Intracavitary applicators

AbstractBackgroundFormulating a clinically acceptable plan within the time‐constrained clinical setting of brachytherapy poses challenges to clinicians. Deep learning based dose prediction methods have shown favorable solutions for enhancing efficiency, but development has primarily been on external beam radiation therapy. Thus, there is a need for translation to brachytherapy.PurposeThis study proposes a dose prediction model utilizing an attention‐gating mechanism and a 3D UNET for cervical cancer high‐dose‐rate intracavitary brachytherapy treatment planning with tandem‐and‐ovoid/ring applicators.MethodsA multi‐institutional data set consisting of 77 retrospective clinical brachytherapy plans was utilized in this study. The data were preprocessed and augmented to increase the number of plans to 252. A 3D UNET architecture with attention gates was constructed and trained for mapping the contour information to dose distribution. The trained model was evaluated on a testing data set using various metrics, including dose statistics and dose‐volume indices. We also trained a baseline UNET model for a fair comparison.ResultsThe attention‐gated 3D UNET model exhibited competitive accuracy in predicting dose distributions similar to the ground truth. The average values of the mean absolute errors were 0.46 ± 11.71 Gy (vs. 0.47 ± 9.16 Gy for a baseline UNET) in CTVHR, 0.55 ± 0.67 Gy (vs. 0.70 ± 1.54 Gy for a baseline UNET) in bladder, 0.42 ± 0.46 Gy (vs. 0.49 ± 1.34 Gy for a baseline UNET) in rectum, and 0.31 ± 0.65 Gy (vs. 0.20 ± 3.76 Gy for a baseline UNET) in sigmoid. Our results showed that the mean individual differences in ΔD2cc for bladder, rectum, and sigmoid were 0.38 ± 1.19 (p = 0.50), 0.43 ± 0.71 (p = 0.41), and −0.47 ± 0.79 (p = 0.30) Gy, respectively. Similarly, the mean individual differences in ΔD1cc for bladder, rectum, and sigmoid were 0.09 ± 1.21 (p = 0.36), 0.20 ± 0.95 (p = 0.24), and −0.21 ± 0.59 (p = 0.30) Gy. The mean individual differences for ΔD90, ΔV100%, ΔV150%, and ΔV200% of the CTVHR were −0.45 ± 2.42 (p = 0.26) Gy, 0.55 ± 9.42% (p = 0.78), 0.82 ± 4.21% (p = 0.81), and −0.80 ± 10.48% (p = 0.36), respectively. The model requires less than 5 s to predict a full 3D dose distribution for a new patient plan.ConclusionAttention‐gated 3D UNET revealed a promising capability in predicting voxel‐wise dose distributions compared to 3D UNET. This model could be deployed for clinical use to predict 3D dose distributions for near real‐time decision‐making before planning, quality assurance, and guiding future automated planning, making the current workflow more efficient.

Retrospective assessment of HDR brachytherapy dose calculation methods in locally advanced cervical cancer patients: AcurosBV vs. AAPM TG43 formalism

AbstractPurposeThis retrospective analysis was completed to investigate the use of a model‐based dose calculation algorithm (MBDCA) AcurosBV, for use in HDR BT treatments for locally advanced cervical cancer treated with tandem and ovoid applicators with interstitial needles.MethodsA cohort of 32 patients receiving post‐EBRT HDR brachytherapy boost with a prescription dose of 5.5 Gy × 5 fractions to the high‐risk clinical target volume (CTVHR) were selected for this study. For standard TG43 dose calculation, applicators were manually digitized on the planning images, while for AcurosBV calculations, solid renderings of Titanium Fletcher Suite Delclos (FSD) applicators included in BrachyVision were matched to those used clinically and Ti needles were manually digitized. The dose was recalculated using Varian's AcurosBV 13.5 and dose‐to‐medium‐in‐medium (Dm,m) was reported. EQD2 values for targets and organs at risk were compared between dose calculation formalisms. D90% and D98% values were reported for the high and intermediate‐risk CTVs, and values were reported for OARs including bladder, rectum, sigmoid, bowel, and vagina. Due to variability within the patient cohort, the dosimetric impact of AcurosBV was investigated corresponding to planning image modality (CT vs. CBCT), presence of Ti needles, and contrast within vaginal balloons used to stabilize implants. AcurosBV showed lower dosimetric values for all plans compared to TG43.ResultsThe average ± standard deviation of dosimetric reduction in D90% was 4.33 ± 0.09% for CTVHR and 4.12 ± 0.09% for CTVIR. The reduction to OARs was: 4.99 ± 0.15% for bladder, 7.87 ± 0.16% for rectum, 5.79 ± 0.17% for sigmoid, 6.91 ± 0.14% for bowel, and 4.55 ± 0.14% for vagina.ConclusionsAcurosBV should be utilized for HDR BT GYN cases, treated with tandem and ovoid applicators, with high degrees of heterogeneity and calculated in tandem with TG43.

Evaluation of cumulative dose distributions from external beam radiation therapy using CT‐to‐CBCT deformable image registration (DIR) for cervical cancer patients

AbstractPurposeTo investigate dose differences between the planning CT (pCT) and dose calculated on pre‐treatment verification CBCTs using DIR and dose summation for cervical cancer patients.MethodsCervical cancer patients treated at our institution with 45 Gy EBRT undergo a pCT and 5 CBCTs, once every five fractions of treatment. A free‐form intensity‐based DIR in MIM was performed between the pCT and each CBCT using the “Merged CBCT” feature to generate an extended FOV‐CBCT (mCBCT). DIR‐generated bladder and rectum contours were adjusted by a physician, and dice similarity coefficients (DSC) were calculated. After deformation, the investigated doses were (1) recalculated in Eclipse using original plan parameters (ecD), and (2) deformed from planning dose (pD) using the deformation matrix in MIM (mdD). Dose summation was performed to the first week's mCBCT. Dose distributions were compared for the bladder, rectum, and PTV in terms of percent dose difference, dose volume histograms (DVHs), and gamma analysis between the calculated doses.ResultsFor the 20 patients, the mean DSC was 0.68 ± 0.17 for bladder and 0.79 ± 0.09 for rectum. Most patients were within 5% of pD for D2cc (19/20), Dmax (17/20), and Dmean (16/20). All patients demonstrated a percent difference > 5% for bladder V45 due to variations in bladder volume from the pCT. D90 showed fewer differences with 19/20 patients within 2% of pD. Gamma rates between pD and ecD averaged 94% for bladder and 94% for rectum, while pD and mdD exhibited slightly better performance for bladder (93%) and lower for rectum (85%).ConclusionUsing DIR with weekly CBCT images, the MIM deformed dose (mdD) was found to be in close agreement with the Eclipse calculated dose (ecD). The proposed workflow should be used on a case‐by‐case basis when the weekly CBCT shows marked difference in organs‐at‐risk from the planning CT.

Setup errors analysis in iterative kV CBCT: A clinical study of cervical cancer treated with Volumetric Modulated Arc Therapy

AbstractObjectiveThis study aims to analyze setup errors in pelvic Volumetric Modulated Arc Therapy (VMAT) for patients with non‐surgical primary cervical cancer, utilizing the onboard iterative kV cone beam CT (iCBCT) imaging system on the Varian Halcyon 2.0 ring gantry structure accelerator to enhance radiotherapy precision.MethodWe selected 132 cervical cancer patients who underwent VMAT with daily iCBCT imaging guidance. Before each treatment session, a registration method based on the bony structure was employed to acquire iCBCT images with the corresponding planning CT images. Following verification and adjustment of image registration results along the three axes (but not rotational), setup errors in the lateral (X‐axis), longitudinal (Y‐axis), and vertical (Z‐axis) directions were recorded for each patient. Subsequently, we analyzed 3642 iCBCT image setup errors.ResultsThe mean setup errors for the X, Y, and Z axes were 4.50 ± 3.79 mm, 6.08 ± 6.30 mm, and 1.48 ± 2.23 mm, respectively. Before correction with iCBCT, setup margins based on the Van Herk formula for the X, Y, and Z axes were 6.28, 12.52, and 3.26 mm, respectively. In individuals aged 60 years and older, setup errors in the X and Y axes were significantly larger than those in the younger group (p < 0.05). Additionally, there is no significant linear correlation between setup errors and treatment fraction numbers.ConclusionData analysis underscores the importance of precise Y‐axis setup for cervical cancer patients undergoing VMAT. Radiotherapy centers without daily iCBCT should appropriately extend the planning target volume (PTV) along the Y‐axis for cervical cancer patients receiving pelvic VMAT. Elderly patients exhibit significantly larger setup errors compared to younger counterparts. In conclusion, iCBCT‐guided radiotherapy is recommended for cervical cancer patients undergoing VMAT to improve setup precision.

Adaptive assessment based on fractional CBCT images for cervical cancer

AbstractPurposeAnatomical and other changes during radiotherapy will cause inaccuracy of dose distributions, therefore the expectation for online adaptive radiation therapy (ART) is high in effectively reducing uncertainties due to intra‐variation. However, ART requires extensive time and effort. This study investigated an adaptive assessment workflow based on fractional cone‐beam computed tomography (CBCT) images.MethodsImage registration, synthetic CT (sCT) generation, auto‐segmentation, and dose calculation were implemented and integrated into ArcherQA Adaptive Check. The rigid registration was based on ITK open source. The deformable image registration (DIR) method was based on a 3D multistage registration network, and the sCT generation method was performed based on a 2D cycle‐consistent adversarial network (CycleGAN). The auto‐segmentation of organs at risk (OARs) on sCT images was finished by a deep learning‐based auto‐segmentation software, DeepViewer. The contours of targets were obtained by the structure‐guided registration. Finally, the dose calculation was based on a GPU‐based Monte Carlo (MC) dose code, ArcherQA.ResultsThe dice similarity coefficient (DSCs) were over 0.86 for target volumes and over 0.79 for OARs. The gamma pass rate of ArcherQA versus Eclipse treatment planning system was more than 99% at the 2%/2 mm criterion with a low‐dose threshold of 10%. The time for the whole process was less than 3 min. The dosimetric results of ArcherQA Adaptive Check were consistent with the Ethos scheduled plan, which can effectively identify the fractions that need the implementation of the Ethos adaptive plan.ConclusionThis study integrated AI‐based technologies and GPU‐based MC technology to evaluate the dose distributions using fractional CBCT images, demonstrating remarkably high efficiency and precision to support future ART processes.

Deep learning‐based auto‐contouring for organs at risk in three‐dimensional image‐guided brachytherapy for cervical cancer and endometrial cancer

Abstract Background Automatic contouring can reduce the time required for delineating organs at risk (OARs) in brachytherapy planning and minimize interobserver variability. Purpose This study aimed to develop and evaluate a deep learning‐based automatic contouring model for OARs in three‐dimensional image‐guided brachytherapy (3D‐IGBT) for cervical and endometrial cancer, including cases with interstitial needles. Methods The dataset comprised 100 patients (140 cases) with cervical or endometrial cancer who underwent 3D‐IGBT. Interstitial needles were used in 74 cases. The nnU‐Net model was trained (80 patients, 80 cases) and tested (20 patients, 60 cases). The OARs considered were the bladder, small bowel, rectum, and sigmoid. Ground truth (GT) contours were manually delineated by radiation oncologists and medical physicists. Evaluation included measuring inference time and assessing geometric agreement using dice similarity coefficient (DSC), surface DSC (sDSC), Hausdorff distance (HD), and 95th percentile HD (95HD). These metrics were also calculated for a combined structure of the rectum and sigmoid (Rec+Sig). Furthermore, D 2cc was calculated based on both the GT and predicted contours using the clinical dose distribution, and the difference between them (ΔD 2cc ) was evaluated. Differences in accuracy with or without interstitial needles were compared using Welch's t ‐test (significance level: p  < 0.05). Results Mean processing time was 30.3 s per case. Mean DSC values for the bladder, small bowel, rectum, sigmoid, and Rec+Sig were 0.96, 0.79, 0.83, 0.76, and 0.87, respectively. Mean 95HD values (mm) were 4.01, 18.8, 13.6, 25.5, and 17.8, respectively; ΔD 2cc values (Gy) were 0.17, 0.53, 0.014, −0.073, and −0.045, respectively. No significant accuracy differences related to interstitial needles were observed for any of the OARs. Conclusions The proposed deep learning model demonstrated potential for application in cases involving interstitial needles and may contribute to improving the efficiency of the treatment planning workflow.

Ranking attention multiple instance learning for lymph node metastasis prediction on multicenter cervical cancer MRI

AbstractPurposeIn the current clinical diagnostic process, the gold standard for lymph node metastasis (LNM) diagnosis is histopathological examination following surgical lymphadenectomy. Developing a non‐invasive and preoperative method for predicting LNM is necessary and holds significant clinical importance.MethodsWe develop a ranking attention multiple instance learning (RA‐MIL) model that integrates convolutional neural networks (CNNs) and ranking attention pooling to diagnose LNM from T2 MRI. Our RA‐MIL model applies the CNNs to derive imaging features from 2D MRI slices and employs ranking attention pooling to create patient‐level feature representation for diagnostic classification. Based on the MIL and attention theory, informative regions of top‐ranking MRI slices from LNM‐positive patients are visualized to enhance the interpretability of automatic LNM prediction. This retrospective study collected 300 female patients with cervical cancer who underwent T2‐weighted magnetic resonance imaging (MRI) scanning and histopathological diagnosis from one hospital (289 patients) and one open‐source dataset (11 patients).ResultsOur RA‐MIL model delivers promising LNM prediction performance, achieving the area under the receiver operating characteristic curve (AUC) of 0.809 on the internal test set and 0.833 on the public dataset. Experiments show significant improvements in LNM status prediction using the proposed RA‐MIL model compared with other state‐of‐the‐art (SOTA) comparative deep learning models.ConclusionsThe developed RA‐MIL model has the potential to serve as a non‐invasive auxiliary tool for preoperative LNM prediction, offering visual interpretability regarding informative MRI slices and regions in LNM‐positive patients.

Inter‐observer variability in library plan selection on iterative CBCT and synthetic CT images of cervical cancer patients

AbstractIntroductionIn the Library‐of‐Plans (LoP) approach, correct plan selection is essential for delivering radiotherapy treatment accurately. However, poor image quality of the cone‐beam computed tomography (CBCT) may introduce inter‐observer variability and thereby hamper accurate plan selection. In this study, we investigated whether new techniques to improve the CBCT image quality and improve consistency in plan selection, affects the accuracy of LoP selection in cervical cancer patients.Materials and methodsCBCT images of 12 patients were used to investigate the inter‐observer variability of plan selection based on different CBCT image types. Six observers were asked to individually select a plan based on clinical X‐ray Volumetric Imaging (XVI) CBCT, iterative reconstructed CBCT (iCBCT) and synthetic CTs (sCT). Selections were performed before and after a consensus meeting with the entire group, in which guidelines were created. A scoring by all observers on the image quality and plan selection procedure was also included. For plan selection, Fleiss' kappa (κ) statistical test was used to determine the inter‐observer variability within one image type.ResultsThe agreement between observers was significantly higher on sCT compared to CBCT. The consensus meeting improved the duration and inter‐observer variability. In this manuscript, the guidelines attributed the overall results in the plan selection. Before the meeting, the gold standard was selected in 76% of the cases on XVI CBCT, 74% on iCBCT, and 76% on sCT. After the meeting, the gold standard was selected in 83% of the cases on XVI CBCT, 81% on iCBCT, and 90% on sCT.ConclusionThe use of sCTs can increase the agreement of plan selection among observers and the gold standard was indicated to be selected more often. It is important that clear guidelines for plan selection are implemented in order to benefit from the increased image quality, accurate selection, and decrease inter‐observer variability.

Identifying the optimal deep learning architecture and parameters for automatic beam aperture definition in 3D radiotherapy

AbstractPurposeTwo‐dimensional radiotherapy is often used to treat cervical cancer in low‐ and middle‐income countries, but treatment planning can be challenging and time‐consuming. Neural networks offer the potential to greatly decrease planning time through automation, but the impact of the wide range of hyperparameters to be set during training on model accuracy has not been exhaustively investigated. In the current study, we evaluated the effect of several convolutional neural network architectures and hyperparameters on 2D radiotherapy treatment field delineation.MethodsSix commonly used deep learning architectures were trained to delineate four‐field box apertures on digitally reconstructed radiographs for cervical cancer radiotherapy. A comprehensive search of optimal hyperparameters for all models was conducted by varying the initial learning rate, image normalization methods, and (when appropriate) convolutional kernel size, the number of learnable parameters via network depth and the number of feature maps per convolution, and nonlinear activation functions. This yielded over 1700 unique models, which were all trained until performance converged and then tested on a separate dataset.ResultsOf all hyperparameters, the choice of initial learning rate was most consistently significant for improved performance on the test set, with all top‐performing models using learning rates of 0.0001. The optimal image normalization was not consistent across architectures. High overlap (mean Dice similarity coefficient = 0.98) and surface distance agreement (mean surface distance < 2 mm) were achieved between the treatment field apertures for all architectures using the identified best hyperparameters. Overlap Dice similarity coefficient (DSC) and distance metrics (mean surface distance and Hausdorff distance) indicated that DeepLabv3+ and D‐LinkNet architectures were least sensitive to initial hyperparameter selection.ConclusionDeepLabv3+ and D‐LinkNet are most robust to initial hyperparameter selection. Learning rate, nonlinear activation function, and kernel size are also important hyperparameters for improving performance.

Three‐dimensional assessment of interfractional cervical and uterine motions using daily magnetic resonance images to determine margins and timing of replanning

AbstractPurposeThis study was conducted to determine the margins and timing of replanning by assessing the daily interfractional cervical and uterine motions using magnetic resonance (MR) images.MethodsEleven patients with cervical cancer, who underwent intensity‐modulated radiotherapy (IMRT) in 23–25 fractions, were considered in this study. The daily and reference MR images were converted into three‐dimensional (3D) shape models. Patient‐specific anisotropic margins were calculated from the proximal 95% of vertices located outside the surface of the reference model. Population‐based margins were defined as the 90th percentile values of the patient‐specific margins. The expanded volume of interest (expVOI) for the cervix and uterus was generated by expanding the reference model based on the population‐based margin to calculate the coverage for daily deformable mesh models. For comparison, expVOIconv was generated using conventional margins: right (R), left (L), anterior (A), posterior (P), superior (S), and inferior (I) were (5, 5, 15, 15, 10, 10) and (10, 10, 20, 20, 15, 15) mm for the cervix and uterus, respectively. Subsequently, a replanning scenario was developed based on the cervical volume change. ExpVOIini and expVOIreplan were generated before and after replanning, respectively.ResultsPopulation‐based margins were (R, L, A, P, S, I) of (7, 7, 11, 6, 11, 8) and (14, 13, 27, 19, 15, 21) mm for the cervix and uterus, respectively. The timing of replanning was found to be the 16th fraction, and the volume of expVOIreplan decreased by >30% compared to that of expVOIini. However, margins cannot be reduced to ensure equivalent coverage after replanning.ConclusionWe determined the margins and timing of replanning through detailed daily analysis. The margins of the cervix were smaller than conventional margins in some directions, while the margins of the uterus were larger in almost all directions. A margin equivalent to that at the initial planning was required for replanning.

Automatic segmentation and applicator reconstruction for CT‐based brachytherapy of cervical cancer using 3D convolutional neural networks

AbstractIn this study, we present deep learning‐based approaches to automatic segmentation and applicator reconstruction with high accuracy and efficiency in the planning computed tomography (CT) for cervical cancer brachytherapy (BT). A novel three‐dimensional (3D) convolutional neural network (CNN) architecture was proposed and referred to as DSD‐UNET. The dataset of 91 patients received CT‐based BT of cervical cancer was used to train and test DSD‐UNET model for auto‐segmentation of high‐risk clinical target volume (HR‐CTV) and organs at risk (OARs). Automatic applicator reconstruction was achieved with DSD‐UNET‐based segmentation of applicator components followed by 3D skeletonization and polynomial curve fitting. Digitization of the channel paths for tandem and ovoid applicator in the planning CT was evaluated utilizing the data from 32 patients. Dice similarity coefficient (DSC), Jaccard Index (JI), and Hausdorff distance (HD) were used to quantitatively evaluate the accuracy. The segmentation performance of DSD‐UNET was compared with that of 3D U‐Net. Results showed that DSD‐UNET method outperformed 3D U‐Net on segmentations of all the structures. The mean DSC values of DSD‐UNET method were 86.9%, 82.9%, and 82.1% for bladder, HR‐CTV, and rectum, respectively. For the performance of automatic applicator reconstruction, outstanding segmentation accuracy was first achieved for the intrauterine and ovoid tubes (average DSC value of 92.1%, average HD value of 2.3 mm). Finally, HDs between the channel paths determined automatically and manually were 0.88 ± 0.12 mm, 0.95 ± 0.16 mm, and 0.96 ± 0.15 mm for the intrauterine, left ovoid, and right ovoid tubes, respectively. The proposed DSD‐UNET method outperformed the 3D U‐Net and could segment HR‐CTV, bladder, and rectum with relatively good accuracy. Accurate digitization of the channel paths could be achieved with the DSD‐UNET‐based method. The proposed approaches could be useful to improve the efficiency and consistency of treatment planning for cervical cancer BT.

Open‐source deep‐learning models for segmentation of normal structures for prostatic and gynecological high‐dose‐rate brachytherapy: Comparison of architectures

AbstractBackgroundThe use of deep learning‐based auto‐contouring algorithms in various treatment planning services is increasingly common. There is a notable deficit of commercially or publicly available models trained on large or diverse datasets containing high‐dose‐rate (HDR) brachytherapy treatment scans, leading to poor performance on images that include HDR implants.PurposeTo implement and evaluate automatic organs‐at‐risk (OARs) segmentation models for use in prostatic‐and‐gynecological computed tomography (CT)‐guided high‐dose‐rate brachytherapy treatment planning.Methods and materials1316 computed tomography (CT) scans and corresponding segmentation files from 1105 prostatic‐or‐gynecological HDR patients treated at our institution from 2017 to 2024 were used for model training. Data sources comprised six CT scanners including a mobile CT unit with previously reported susceptibility to image streaking artifacts. Two UNet‐derived model architectures, UNet++ and nnU‐Net, were investigated for bladder and rectum model training. The models were tested on 100 CT scans and clinically‐used segmentation files from 62 prostatic‐or‐gynecological HDR brachytherapy patients, disjoint from the training set, collected in 2024. Performance was evaluated using the Dice‐Similarity‐Coefficient (DSC) between model predicted contours and clinically‐used contours on slices in common with the Clinical‐Target‐Volume (CTV). Additionally, a blinded evaluation of ten random test cases was conducted by three experienced planners.ResultsMedian (interquartile range) 3D DSC on CTV‐containing slices were 0.95 (0.04) and 0.87 (0.09) for the UNet++ bladder and rectum models, respectively, and 0.96 (0.03) and 0.88 (0.10) for the nnU‐Net. The rank‐sum test did not reveal statistically significant differences in these DSC (p = 0.15 and 0.27, respectively). The blinded evaluation scored trained models higher than clinically‐used contours.ConclusionBoth UNet‐derived architectures perform similarly on the bladder and rectum and are adequately accurate to reduce contouring time in a review‐and‐edit context during HDR brachytherapy planning. The UNet++ models were chosen for implementation at our institution due to lower computing hardware requirements and are in routine clinical use.

Commissioning of the Varian universal interstitial cylinder system for HDR brachytherapy of gynecological cancer

AbstractPurposeThis paper outlines the commissioning of the Varian (VMS, Varian Medical Systems, Palo Alto, CA) Universal Interstitial Cylinder (UIC) applicator set for Ir‐192 HDR brachytherapy. The UIC was commissioned for use with CT and MRI and a custom phantom was designed to avoid the introduction of water‐like materials into the needle guide tracks. Various marker strands were investigated to determine which allowed the most accurate reconstruction of source positions.MethodsPlanar kV and MV imaging, along with physical measurements and autoradiographs, were used to commission the physical dimensions of all components of the UIC applicator set. CT and MR imaging was used to further commission one configuration of the UIC with UCP and eight interstitial needles in a simulated clinical setup using a GYN phantom. Three different methods of channel identification were compared – no radio opaque markers, VMS numbered markers, or nylon coated stainless steel leader wires – to see which best aided in channel identification and image registration. An HDR MRI Lumen marker (C4 Imaging, LLC) was used to verify any applicator rotation on MR scans during image registration. Three types of GYN phantoms were investigated – wet towel, gelatin, and ground beef. Dimensions of all components were compared with vendor provided information, including the solid applicator models, which are based on the computer‐aided design model files of the specific applicators.ResultsThe dimensions of the applicators could be validated using physical measurements, kV and MV planar imaging, and CT scans. The ground beef based GYN phantom best eliminated the introduction of water into the needle guide tracks that was found when using a water or gel‐based phantom. CT scans using no radio opaque markers did not allow the plastic needles to be visualized well enough to digitize source positions. CT scans with VMS markers showed significant artifact. CT scans with the nylon coated stainless steel wires provided the best visibility of the needle locations to aid in digitizing source positions. The use of an MR marker allowed the channel to be identified on the MR scan and confirm rotation for image registration.ConclusionsThe UIC set and applicator configuration was commissioned for CT and MR based treatment planning. The plastic components of the UIC applicator set pose challenges to the commissioning process but the use of radio opaque markers seen on CT combined with MR image registration allow the source positions within the needles, as well as the location of the end of the needles, to be digitized appropriately. A ground beef phantom minimized the fluid introduced into the needle guide track, minimizing any unintended MR and CT signal in the needle guide tracks.

Development and performance assessment of an advanced Lucas‐Kanade algorithm for dose mapping of cervical cancer external radiotherapy and brachytherapy plans

AbstractPurposeThe aim of this study was to verify the possibility of summing the dose distributions of combined radiotherapeutic treatment of cervical cancer using the extended Lucas‐Kanade algorithm for deformable image registration.Materials and methodsFirst, a deformable registration of planning computed tomography images for the external radiotherapy and brachytherapy treatment of 10 patients with different parameter settings of the Lucas‐Kanade algorithm was performed. By evaluating the registered data using landmarks distance, root mean square error of Hounsfield units and 2D gamma analysis, the optimal parameter values were found. Next, with another group of 10 patients, the accuracy of the dose mapping of the optimized Lucas‐Kanade algorithm was assessed and compared with Horn‐Schunck and modified Demons algorithms using dose differences at landmarks.ResultsThe best results of the Lucas‐Kanade deformable registration were achieved for two pyramid levels in combination with a window size of 3 voxels. With this registration setting, the average landmarks distance was 2.35 mm, the RMSE was the smallest and the average gamma score reached a value of 86.7%. The mean dose difference at the landmarks after mapping the external radiotherapy and brachytherapy dose distributions was 1.33 Gy. A statistically significant difference was observed on comparing the Lucas‐Kanade method with the Horn‐Schunck and Demons algorithms, where after the deformable registration, the average difference in dose was 1.60 Gy (P‐value: 0.0055) and 1.69 Gy (P‐value: 0.0012), respectively.ConclusionLucas‐Kanade deformable registration can lead to a more accurate model of dose accumulation and provide a more realistic idea of the dose distribution.

Modeling dosimetric benefits from daily adaptive RT for gynecological cancer patients with and without knowledge‐based dose prediction

AbstractPurposeDaily online adaptive radiotherapy (ART) improves dose metrics for gynecological cancer patients, but the on‐treatment process is resource‐intensive requiring longer appointments and additional time from the entire adaptive team. To optimize resource allocation, we propose a model to identify high‐priority patients.MethodsFor 49 retrospective cervical and endometrial cancer patients, we calculated two initial plans: the treated standard‐of‐care (InitialSOC) and a reduced margin initial plan (InitialART) for adapting with the Ethos treatment planning system. Daily doses corresponding to standard and reduced margins (DailySOC and DailyART) were determined by re‐segmenting the anatomy based on the treatment CBCT and calculating dose on a synthetic CT. These initial and daily doses were used to estimate the ART benefit (= DailySOC‐DailyART) versus initial plan differences (= InitialSOC–InitialART) via multivariate linear regression. Dosimetric benefits were modeled with initial plan differences () of (cc), (Gy), and (Gy). Anatomy (intact uterus or post‐hysterectomy), DoseType (simultaneous integrated boost [SIB] vs. single dose), and/or prescription value. To establish a logistic model, we classified the top 10% in each metric as high‐benefit patients. We then built a logistic model to predict these patients from the previous predictors. Leave‐one‐out validation and ROC analysis were used to evaluate the accuracy. To improve the clinical efficiency of this predictive process, we also created knowledge‐based plans for the ΔInitial plans () and repeated the analysis.ResultsIn both and our multivariate analysis showed low R2 values 0.34–0.52 versus 0.14–0.38. The most significant predictor in each multivariate model was the corresponding ∆Initial metric (e.g., Bowel (V40 Gy), p < 1e−05). In the logistic model, the metrics with the strongest correlation to the high‐benefit patients were (cc), (Gy), , and prescription. The models for original and knowledge‐based plans had an AUC of 0.85 versus 0.78. The sensitivity and specificity were 0.92/0.72 and 0.69/0.80, respectively.ConclusionThis methodology will allow clinics to prioritize patients for resource‐intensive daily online ART.

Impact of rectal gas evacuation on dosimetry and applicator displacement in cervical cancer brachytherapy

Abstract Objective This study aimed to evaluate the impact of rectal gas evacuation on organ‐at‐risk (OAR) volumes, dose‐volume histogram (DVH) parameters, and applicator displacement during cervical cancer brachytherapy. Methods Twenty‐one cervical cancer patients who received three‐dimensional brachytherapy at our center between November and December 2024 and presented with rectal gas were retrospectively included. Planning computed tomography (CT) images were acquired before and after rectal gas evacuation to evaluate changes in rectal and bladder volumes, as well as radiation dose variations to OARs (bladder, rectum, sigmoid, and small intestine). Dosimetric parameters analyzed comprised D 0.1cc , D 1cc , D 2cc , and D 5cc (minimum doses delivered to the most irradiated 0.1, 1, 2, and 5 cm 3 of the OARs, respectively), as well as D max (maximum dose) and D mean (mean dose). Displacements of the applicator tip and cervical stopper were quantified using a coordinate system based on pelvic bony landmarks. Results Rectal volume decreased by 40.1% after gas evacuation, while bladder volume increased by 18.2%. D 0.1cc , D 1cc , D 2cc , D 5cc , and D max in the rectum decreased significantly ( P  < 0.001) after gas evacuation, whereas no significant changes were observed in the DVH parameters of the other OARs. The mean displacements of the applicator tip and cervical stopper were 5.86 ± 3.64 mm and 4.23 ± 3.30 mm, respectively. Conclusions Rectal gas evacuation results in a statistically significant and clinically relevant reduction in rectal volume and rectal dose, underscoring its importance as a routine clinical procedure. However, as it may induce millimeter‐level applicator displacement with clinically measurable dosimetric consequences, careful monitoring is warranted, with post‐evacuation replanning or, if necessary, applicator adjustment.

Cumulative dose assessment with transformer‐based deformable image registration addition for cervical cancer patients

Abstract Purpose Based on the combination mode of internal and external radiotherapy for cervical cancer, this study aimed to investigate an accurate transformer‐based deformable image registration (DIR) method for cumulative dose assessment. Methods and Materials According to a retrospective analysis conducted on 180 patients with cervical cancer who underwent intracavitary brachytherapy (ICBT) and external beam radiation therapy (EBRT), this study proposed a mix‐transformer structure‐based deformable image registration (MTDIR) network for registering CT scans of ICBT and EBRT, followed by dose accumulation and assessment. The mean dice similarity coefficient (DSC) and Hausdorff distance (HD) of the rectum and bladder were computed to compare the performance of MTDIR with that of the state‐of‐the‐art VoxelMorph method and the DIR method provided by velocity. Additionally, the cumulative dose of the bladder and rectum from four methods was calculated: direct DVH parameter addition (DA), DIR‐based dose addition provided by Velocity (VA), VoxelMorph‐based dose addition (VoA), and MTDIR‐based dose addition (MA). Results The mean DSC values of MTDIR, VoxelMorph, and Velocity for the bladder were 0.78, 0.75, and 0.72 for the registration between CT scans of EBRT and the last ICBT, respectively. The mean DSC values of the rectum were also equal to 0.58, 0.56, and 0.52. The mean HDmean values of MTDIR, VoxelMorph, and Velocity for the bladder were 4.81, 5.16, and 5.43, and the mean HDmean values of the rectum were 5.41, 5.81, and 6.93, respectively. For the registration between CT scans of ICBT, the mean DSC values of MTDIR, VoxelMorph, and Velocity for the bladder were obtained 0.82, 0.80, and 0.77, and the mean DSC values for the rectum were equal to 0.70, 0.68, and 0.63, respectively. The mean HDmean values of MTDIR, VoxelMorph, and Velocity for the bladder were 4.22, 4.58, and 4.79, and the mean HDmean values of the rectum were obtained 4.86, 5.17, and 5.64, respectively. The results generally suggested that MTDIR outperformed VoxelMorph and Velocity. Conclusions The study findings demonstrated that the model developed based on parameters obtained from the proposed method exhibited higher registration accuracy.

Prospective multi‐institutional study of library‐based adaptive radiotherapy for cervical cancer: Evaluation of plan‐of‐the‐day selection and population analysis

Abstract Purpose Plan‐of‐the‐day (PoD) adaptive radiation therapy (ART) is based on a library of treatment plans, with 3D daily imaging guiding the plan selection. In a phase II multi‐institutional trial of cone‐beam CT (CBCT)‐guided PoD‐ART for locally advanced cervical carcinoma (LACC), this study aimed at evaluating the PoD selection, its geometric and dosimetric impact and characterizing a sub‐population of patients associated with dosimetric improvement from ART. Material and methods For 49 cervical cancer patients, three planning CT scans [empty bladder (EB), intermediate bladder (IB) and full bladder (FB)] were acquired to generate a treatment plan library. A dose of 45 Gy was prescribed to the planning target volume in 25 fractions. Daily CBCT were acquired to visually select the best plan in the library (Manual‐ART strategy). A deep learning model was used to segment daily clinical target volume (CTVt) and organs‐at‐risk (OAR). Manual‐ART was compared to two strategies: (i) “Non‐ART” strategy (IB‐CT treatment plan only); (ii) PoD‐ART strategy selecting the PoD maximizing CTVt coverage (“Cov‐ART”). Geometrical and dosimetric coverages of daily CTVt and OAR were assessed. Decision trees were developed to predict the subpopulation of patients associated with dosimetric benefit from PoD‐ART. Results The agreement in PoD selection between Manual‐ART and Cov‐ART was 63.5%. Compared to the Non‐ART strategy (D95%‐CTV: 43.6 ± 4.1 Gy), PoD‐ART significantly increased the dose to the target, with Manual‐ART achieving 44.0 ± 3.0 Gy and Cov‐ART with 44.1 ± 2.0 Gy. Decision trees using IB‐CT plan and first two treatment fractions correctly classified 85.4% and 93.8% of patients as benefiting or not from PoD‐ART. Conclusions In PoD‐ART for LACC, selected treatment plans by the radiation oncologist had 63.5% concordance with treatment plans maximizing target coverage. PoD‐ART increased dose to target, without compromising dose to OARs, with the largest benefit observed in a sub‐population identifiable after two treatment fractions.

A prior‐information‐based automatic segmentation method for the clinical target volume in adaptive radiotherapy of cervical cancer

AbstractObjectiveAdaptive planning to accommodate anatomic changes during treatment often requires repeated segmentation. In this study, prior patient‐specific data was integrateda into a registration‐guided multi‐channel multi‐path (Rg‐MCMP) segmentation framework to improve the accuracy of repeated clinical target volume (CTV) segmentation.MethodsThis study was based on CT image datasets for a total of 90 cervical cancer patients who received two courses of radiotherapy. A total of 15 patients were selected randomly as the test set. In the Rg‐MCMP segmentation framework, the first‐course CT images (CT1) were registered to second‐course CT images (CT2) to yield aligned CT images (aCT1), and the CTV in the first course (CTV1) was propagated to yield aligned CTV contours (aCTV1). Then, aCT1, aCTV1, and CT2 were combined as the inputs for 3D U‐Net consisting of a channel‐based multi‐path feature extraction network. The performance of the Rg‐MCMP segmentation framework was evaluated and compared with the single‐channel single‐path model (SCSP), the standalone registration methods, and the registration‐guided multi‐channel single‐path (Rg‐MCSP) model. The Dice similarity coefficient (DSC), 95% Hausdorff distance (HD95), and average surface distance (ASD) were used as the metrics.ResultsThe average DSC of CTV for the deformable image DIR‐MCMP model was found to be 0.892, greater than that of the standalone DIR (0.856), SCSP (0.837), and DIR‐MCSP (0.877), which were improvements of 4.2%, 6.6%, and 1.7%, respectively. Similarly, the rigid body DIR‐MCMP model yielded an average DSC of 0.875, which exceeded standalone RB (0.787), SCSP (0.837), and registration‐guided multi‐channel single‐path (0.848), which were improvements of 11.2%, 4.5%, and 3.2%, respectively. These improvements in DSC were statistically significant (p < 0.05).ConclusionThe proposed Rg‐MCMP framework achieved excellent accuracy in CTV segmentation as part of the adaptive radiotherapy workflow.

Position‐dependent offset corrections for ring applicator reconstruction in cervical cancer brachytherapy

AbstractPurposeDue to the tight curvature in their design, ring applicators are usually associated with large positioning errors. The standard practice to correct for these deviations based on global offsets may not be sufficient to comply with the recommended tolerance. In this work, we investigate two methods for applicator reconstruction that implement position‐dependent source offset corrections.MethodsMeasurements were performed using the Varian Interstitial PEEK Ring 60° and a Varian BRAVOS afterloader. Source positioning was characterized by means of autoradiographs acquired for three different loading patterns and three 192Ir sources over a period of 5 months. Additionally, the actual source path was determined by means of a series of planar kV images for different dummy cable positions. The first position‐dependent correction method consists of locally modifying the radius of the reconstructed source path according to the measured offsets. The second method, recommended by Varian, simulates a bidirectional movement of the source during applicator reconstruction to compensate for positioning errors.ResultsAutoradiographs showed a quasi‐linear increase of the dwell position offsets, with a negligible error at the tip and a value close to 3 mm at the end of the ring. This result, consistent with a circular wire movement with an effective radius 0.5 mm larger than the nominal value, was in agreement with the observations from the kV images. After implementation of the position‐dependent correction methods, residual positioning errors for the two methods resulted in a mean value (±1 SD) of 0.0 (±0.3) mm, and a range of [−0.7 mm, 0.7 mm].ConclusionThe two tested methods for applicator reconstruction with position‐dependent source offset corrections were able to successfully correct the positioning errors. The method recommended by the manufacturer had the additional advantages of a more straightforward implementation and the potential for use in other applicator types.

Clinical application and improvement of a CNN‐based autosegmentation model for clinical target volumes in cervical cancer radiotherapy

AbstractObjectiveClinical target volume (CTV) autosegmentation for cervical cancer is desirable for radiation therapy. Data heterogeneity and interobserver variability (IOV) limit the clinical adaptability of such methods. The adaptive method is proposed to improve the adaptability of CNN‐based autosegmentation of CTV contours in cervical cancer.MethodsThis study included 400 cervical cancer treatment planning cases with CTV delineated by radiation oncologists from three hospitals. The datasets were divided into five subdatasets (80 cases each). The cases in datasets 1, 2, and 3 were delineated by physicians A, B, and C, respectively. The cases in datasets 4 and 5 were delineated by multiple physicians. Dataset 1 was divided into training (50 cases), validation (10 cases), and testing (20 cases) cohorts, and they were used to construct the pretrained model. Datasets 2–5 were regarded as host datasets to evaluate the accuracy of the pretrained model. In the adaptive process, the pretrained model was fine‐tuned to measure improvements by gradually adding more training cases selected from the host datasets. The accuracy of the autosegmentation model on each host dataset was evaluated using the corresponding test cases. The Dice similarity coefficient (DSC) and 95% Hausdorff distance (HD_95) were used to evaluate the accuracy.ResultsBefore and after adaptive improvements, the average DSC values on the host datasets were 0.818 versus 0.882, 0.763 versus 0.810, 0.727 versus 0.772, and 0.679 versus 0.789, which are improvements of 7.82%, 6.16%, 6.19%, and 16.05%, respectively. The average HD_95 values were 11.143 mm versus 6.853 mm, 22.402 mm versus 14.076 mm, 28.145 mm versus 16.437 mm, and 33.034 mm versus 16.441 mm, which are improvements of 37.94%, 37.17%, 41.60%, and 50.23%, respectively.ConclusionThe proposed method improved the adaptability of the CNN‐based autosegmentation model when applied to host datasets.

Dosimetric impact of edema on inguinal lymph node boost in locally advanced vulvar cancer

AbstractWe aimed to evaluate the extent of groin edema and its dosimetric effect in boosted inguinal lymph nodes (LN) for vulvar cancer patients. The level of edema was determined in 10 patients treated with radical radiotherapy. A dosimetric evaluation of six LNs in the patient with the maximum level of edema was performed. The accumulated dose across CBCT fractions was acceptable for all six LNs (>94% of prescribed dose) even with the development of up to 13 mm of edema. The major contributor to fractional dose degradation was geographical displacement of the nodes. We suggest evaluation of edema on daily CBCT.

Evaluation of bladder filling effects on the dose distribution during radiotherapy for cervical cancer based on daily CT images

AbstractPurposeThis study aimed to assess the effects of bladder filling during cervical cancer radiotherapy on target volume and organs at risk (OARs) dose based on daily computed tomography (daily‐CT) images and provide bladder‐volume‐based dose prediction models.MethodsNineteen patients (475 daily‐CTs) comprised the study group, and five patients comprised the validation set (25 daily‐CTs). Target volumes and OARs were delineated on daily‐CT images and the treatment plan was recalculated accordingly. The deviation from the planning bladder volume (DVB), the correlation between DVB and clinical (CTV)/planning (PTV) target volume in terms of prescribed dose coverage, and the relationship of small bowel volume and bladder dose with the ratio of bladder volume (RVB) were analyzed.ResultsIn all cases, the prescribed dose coverage in the CTV was >95% when DVB was <200 cm3, whereas that in the PTV was >95% when RVB was <160%. The ratio of bladder V45 Gy to the planning bladder V45 Gy (RBV45) exhibited a negative linear relationship with RVB (RBV45 = −0.18*RVB + 120.8; R2 = 0.80). Moreover, the ratio of small bowel volume to planning small bowel volume (RVS) exhibited a negative linear relationship with RVB (RVS = −1.06*RVB +217.59; R2 = 0.41). The validation set results showed that the linear model predicted well the effects of bladder volume changes on target volume coverage and bladder dose.ConclusionsThis study assessed dosimetry and volume effects of bladder filling on target and OARs based on daily‐CT images. We established a quantitative relationship between these parameters, providing dose prediction models for cervical cancer radiotherapy.

Plan quality and treatment efficiency assurance of two VMAT optimization for cervical cancer radiotherapy

AbstractTo investigate the difference of the fluence map optimization (FMO) and Stochastic platform optimization (SPO) algorithm in a newly‐introduced treatment planning system (TPS). Methods: 34 cervical cancer patients with definitive radiation were retrospectively analyzed. Each patient has four plans: FMO with fixed jaw plans (FMO‐FJ) and no fixed jaw plans (FMO‐NFJ); SPO with fixed jaw plans (SPO‐FJ) and no fixed jaw plans (SPO‐NFJ). Dosimetric parameters, Modulation Complexity Score (MCS), Gamma Pass Rate (GPR) and delivery time were analyzed among the four plans. Results: For target coverage, SPO‐FJ plans are the best ones (P ≤ 0.00). FMO plans are better than SPO‐NFJ plans (P ≤ 0.00). For OARs sparing, SPO‐FJ plans are better than FMO plans for mostly OARs (P ≤ 0.04). Additionally, SPO‐FJ plans are better than SPO‐NFJ plans (P ≤ 0.02), except for rectum V45Gy. Compared to SPO‐NFJ plans, the FMO plans delivered less dose to bladder, rectum, colon V40Gy and pelvic bone V40Gy (P ≤ 0.04). Meanwhile, the SPO‐NFJ plans showed superiority in MU, delivery time, MCS and GPR in all plans. In terms of delivery time and MCS, the SPO‐FJ plans are better than FMO plans. FMO‐FJ plans are better than FMO‐NFJ plans in delivery efficiency. MCSs are strongly correlated with PCTV length, which are negatively with PCTV length (P ≤ 0.03). The delivery time and MUs of the four plans are strongly correlated (P ≤ 0.02). Comparing plans with fixed or no fixed jaw in two algorithms, no difference was found in FMO plans in target coverage and minor difference in Kidney_L Dmean, Mu and delivery time between PCTV width≤15.5 cm group and >15.5 cm group. For SPO plans, SPO‐FJ plans showed more superiority in target coverage and OARs sparing than the SPO‐NFJ plans in the two groups. Conclusions: SPO‐FJ plans showed superiority in target coverage and OARs sparing, as well as higher delivery efficiency in the four plans.

Implementation of needle‐tracking technology for real‐time transrectal ultrasound‐guided interstitial gynecological HDR brachytherapy: A feasibility study

AbstractPurposeTo investigate the feasibility of adapting a commercial prostate biopsy system for transrectal ultrasound (TRUS)‐guided hybrid gynecological (GYN) high‐dose‐rate (HDR) brachytherapy (BT). Leveraging 3D‐TRUS and MR image fusion, the prototype system aims to improve real‐time needle placement accuracy.Materials and MethodsA second‐generation, multi‐imaging modality female pelvic phantom was developed to validate the system's feasibility. Software and hardware modifications, including user‐accessible calibration modules and a redesigned needle guide for multi‐needle insertion, were made to the pre‐existing commercial system to enable use for GYN BT applications. An end‐to‐end feasibility test was performed to acquire 3D‐TRUS images, perform contour‐based registration with pre‐implant MR, and insert six needles to targeted locations under real‐time TRUS guidance. A 30° tandem without ovoids was added to mimic a hybrid GYN implant. The most proximal and most distal distances between the planned needle track and the visible portion of each inserted needle were measured. A CT/MR image‐based treatment plan with a prescribed dose of 6 Gy was generated for the resulting 3D‐TRUS‐guided implant (tandem and needles) within the phantom.ResultsThe modified phantom improved needle visualization and insertion range by de‐gassing the silicone and increasing the window size. The system accuracy for average ± standard deviations from intended needle tracks was 1.31 ± 1.36 mm (proximal) and 2.04 ± 2.05 mm (distal). Post‐implant imaging confirmed needle placement and good target coverage. Needle placement was verified on CT/MR images and treatment plan quality was clinically acceptable.ConclusionsWith enhanced needle placement accuracy and integrated clinical workflow, this study demonstrates the feasibility of adapting a commercially available prostate biopsy system for 3D‐TRUS‐guided hybrid GYN HDR BT.

Novel validation of HDR brachytherapy dosimetry for cervical cancer using egs_brachy Monte Carlo simulations: a comparative analysis with Oncentra treatment planning system

Abstract Purpose This study aims to validate HDR brachytherapy dosimetry for cervical cancer patients utilizing the egs_brachy Monte Carlo (MC) simulation. Methods Three cervical cancer patients treated with 192 Ir HDR brachytherapy were included. Dose distributions were calculated by the Oncentra Brachy v4 treatment planning system (TPS) based on AAPM TG‐43. The newly developed eb_gui, an egs_brachy graphical user interface for MC simulations, was applied in recalculating dose distributions for 12 fractions using digital imaging and communications in medicine‐radiotherapy (DICOM‐RT) anatomical information. Comparisons were made for clinical target volume (CTV), bladder, and rectum using dose–volume histograms (DVH) and clinically relevant plan quality indices. Results TPS‐calculated doses were greater than those obtained from MC simulations. For the CTV, the median percentage differences were 7.9% (Q1: 6.4%, Q3: 9.8%; range: 0.4%–10.4%) for D 90 . For the bladder, the median percentage differences were 0.7% (Q1: 0.4%, Q3: 2.3%; range: −9.4–5.4%) for D 2cc . For the rectum, the median percentage differences were 3.6% (Q1: 2.8%, Q3: 5.6%; range: 0.9%–6.4%) for D 2cc . Conclusion CTV and critical organ doses calculated by the TPS were consistently greater than those obtained from MC simulations. This suggests that the TPS may overestimate dose distributions, especially in heterogeneous regions like the pelvis. These results emphasize the need for continued validation of TPS algorithms in HDR brachytherapy for cervical cancer.

Comparison of two inverse planning algorithms for cervical cancer brachytherapy

Abstract Purpose To compare two inverse planning algorithms, the hybrid inverse planning optimization (HIPO) algorithm and the inverse planning simulated annealing (IPSA) algorithm, for cervical cancer brachytherapy and provide suggestions for their usage. Material and methods This study consisted of 24 cervical cancer patients treated with CT image‐based high‐dose‐rate brachytherapy using various combinations of tandem/ovoid applicator and interstitial needles. For fixed catheter configurations, plans were retrospectively optimized with two methods: IPSA and HIPO. The dosimetric parameters with respect to target coverage, localization of high dose volume (LHDV), conformal index (COIN), and sparing of organs at risk (OARs) were evaluated. A plan assessment method which combines a graphical analysis and a scoring index was used to compare the quality of two plans for each case. The characteristics of dwell time distributions of the two plans were also analyzed in detail. Results Both IPSA and HIPO can produce clinically acceptable treatment plans. The rectum D 2cc was slightly lower for HIPO as compared to IPSA ( P  = 0.002). All other dosimetric parameters for targets and OARs were not significantly different between the two algorithms. The generated radar plots and scores intuitively presented the plan properties and enabled to reflect the clinical priorities for the treatment plans. Significant different characteristics were observed between the dwell time distributions generated by IPSA and HIPO. Conclusions Both algorithms could generate high‐quality treatment plans, but their performances were slightly different in terms of each specific patient. The clinical decision on the optimal plan for each patient can be made quickly and consistently with the help of the plan assessment method. Besides, the characteristics of dwell time distribution were suggested to be taken into account during plan selection. Compared to IPSA, the dwell time distributions generated by HIPO may be closer to clinical preference.

The clinical impact of removing rectal gas on high‐dose‐rate brachytherapy dose distributions for gynecologic cancers

AbstractPurposeTo evaluate the impact of gas removal on bladder and rectal doses during intracavitary and interstitial high‐dose‐rate brachytherapy (HDRB) for gynecologic cancers.Material and MethodsFifteen patients treated with definitive external beam radiation followed by HDRB for gynecologic cancers for a total of 21 fractions, presented with a significant amount of rectal gas at initial CT imaging (CTGAS) after implantation. The gas was removed via rectal tubing followed by subsequent scan acquisition (CTCLINICAL), which was used for planning and treatment delivery. To assess the effect of gas removal on dosimetry, both bladder and rectum volumes were recontoured on CTGAS. In order to evaluate the clinical impact on the total Equivalent‐Dose‐in‐2Gy‐fraction (EQD2), each fraction was also replanned to maintain clinically delivered target coverage (HRCTV D90). EQD2 D2cm3 for bladder and rectum were compared between plans. The Wilcoxon signed rank test was performed to evaluate statistically significant differences for all comparisons (P < 0.05).ResultsMean rectum and bladder Dmax, D0.1cm3, D1cm3, D2cm3, and D5cm3 were significantly different between CTGAS and CTCLINICAL. The mean percent increases on CTGAS for bladder were 12.3, 8.4, 9.9, 10.2, and 9.5% respectively and for rectum were 27.0, 19.6, 18.1, 18.5, and 19.4%, respectively. After replanning with CTGAS to maintain HRCTV D90 EQD2, bladder and rectum EQD2 D2 cm3 resulted in significantly higher doses. The mean EQD2 D2 cm3 difference was 2.4 and 4.1 Gy for bladder and rectum, revealing a higher impact of gas removal on rectal DVH.ConclusionRectal gas removal resulted in statistically significant differences for both bladder and rectum. The resulting larger EQD2 D2 cm3 for bladder and rectum demonstrates that if patients were treated without removing gas, target coverage would need to be sacrificed to satisfy the rectum constraints and prevent toxicities. Therefore, this study demonstrates the importance of gas removal for gynecologic HDRB patients.

Characteristics of preplan‐based three‐dimensional individual template‐guided brachytherapy compared to freehand implantation

AbstractImage‐guided adaptive intracavitary/interstitial brachytherapy (IC/IS IGABT) has exhibited superior dosimetry advantage and local control for locally advanced cervical cancer (LACC). Our group designed a type of cylindrical three‐dimensional (3D) printed vaginal template combining an intracavitary applicator with straight and oblique interstitial needles according to the preplan on computed tomography images. This work aimed to research the consistency of the preplan with the treatment plan at every fraction to verify the practical guiding significance of the preplan. We also investigated the difference between 3D‐printed template‐guided implantation compared with freehand implantation for LACC. Twenty‐six patients were treated with 3D‐printed individual templates (3D template group), and 20 patients were treated by using freehand insertion (freehand group). Patients in the 3D template group would take a preplan one week before treatment to design and print the individual template, while the freehand group did not. All patients accepted volumetric rotational intensity‐modulated radiotherapy at a dose of 49.4 Gy in 26 fractions and subsequent brachytherapy at a dose of 26 Gy in four fractions. All analyses were performed by utilizing SPSS 26. The insertion depth was decreased in fractions 1 and 4 compared with the preplan. None of the dose volume histogram parameters of fractions 1–3, nor the D2cc of bladder and bowel at fraction 4 were barely changed compared with the preplan. The D90 and D98 of the high‐risk clinical target volume in the 3D template group were statistically higher than those in the freehand group (p < 0.01). The D2cc of the rectum, bladder, bowel, and sigmoid in the 3D template group were all lower than those in the freehand group (p < 0.01). The preplan in this research is consistent with treatment plans, which is important to ensure the feasibility of applying a 3D‐printed template in brachytherapy. The 3D‐printed individual guidance template was an effective method in brachytherapy for locally advanced cervical cancer.

Dosimetric evaluation of the feasibility of utilizing a reduced number of interstitial needles in combined intracavitary and interstitial brachytherapy for cervical cancer

AbstractPurposeTo evaluate the ability of the Venezia advanced multichannel tandem and ring applicator to consistently produce dosimetrically comparable plans utilizing a reduced number of needle channels, to reduce the risk of secondary complications when boosting cervical cancer treatments with high dose rate (HDR) brachytherapy.MethodsWe evaluated 26 fractions from 13 patients who were treated with HDR brachytherapy using the Venezia (Elekta) applicator. The original plans included a full load of 12–16 needles, including both parallel and 30‐degree oblique needles. We replanned each original to nine new configurations, with a reduced number of two, three, four, or six needles. Comparisons included differences in percentage dose coverage to 90% of the high‐risk clinical target volume, and percentage dose to 2 cm3 of the bladder, rectum, sigmoid, and bowel. We considered new plans “passing” if they remained within our standards (D90 > 100%; D2 cm3 < 85% bladder, <65% rectum, sigmoid, bowel) or did not perform worse than original.ResultsRemoving only the two most anterior or the two most posterior needles from both sides showed 80.8% and 61.5% overall passing rate. Removal of the most anterior and posterior four needles together showed 65.4% overall passing rate. Removing all oblique needles showed 19.2% overall passing rate. Removing only left‐sided or only right‐sided oblique needles showed 46.2% and 23.1% overall passing, respectively. Removing only right‐sided or only left‐sided parallel needles separately showed 19.2% and 34.6% overall passing, respectively. Removing all parallel needles showed 11.5% overall passing rate.ConclusionsAs only two replans required a full needle load to maintain dosimetric quality and 40 (76.9%), 36 (34.6%), 18 (69.2%), and 10 (19.2%) replans passed with 2, 3, 4, and 6 needles removed respectively, this indicates the potential for using a lesser number of interstitial needles during combined intracavitary and interstitial HDR brachytherapy while maintaining dosimetric quality.

Dosimetric and feasibility evaluation of a CBCT‐based daily adaptive radiotherapy protocol for locally advanced cervical cancer

AbstractPurposeEvaluate a cone‐beam computed tomography (CBCT)‐based daily adaptive platform in cervical cancer for multiple endpoints: (1) physics contouring accuracy of daily CTVs, (2) CTV coverage with adapted plans and reduced PTV margins versus non‐adapted plans with standard‐of‐care (SOC) margins, (3) dosimetric improvements to CTV and organs‐at‐risk (OARs), and (4) on‐couch time.Methods and materialsUsing a Varian Ethos™ emulator and KV‐CBCT scans, we simulated the doses 15 retrospective cervical cancer patients would have received with/without online adaptation for five fractions. We compared contours and doses from SOC plans (5–15 mm CTV‐to‐PTV margins) to adapted plans (3 mm margins). Auto‐segmented CTVs and OARs were reviewed and edited by trained physicists. Physics‐edited targets were evaluated by an oncologist. Time spent reviewing and editing auto‐segmented structures was recorded. Metrics from the CTV (D99%), bowel (V45Gy, V40Gy), bladder (D50%), and rectum (D50%) were compared.ResultsThe physician approved the physics‐edited CTVs for 55/75 fractions; 16/75 required reductions, and 4/75 required CTV expansions. CTVs were encapsulated by unadapted, SOC PTVs for 56/75 (72%) fractions—representative of current clinical practice. CTVs were completely covered by adapted 3 mm PTVs for 71/75 (94.6%) fractions. CTV D99% values for adapted plans were comparable to non‐adapted SOC plans (average difference of −0.9%), while all OAR metrics improved with adaptation. Specifically, bowel V45Gy and V40Gy decreased on average by 87.6 and 109.4 cc, while bladder and rectum D50% decreased by 37.7% and 35.8%, respectively. The time required for contouring and calculating an adaptive plan for 65/75 fractions was less than 20 min (range: 1–29 min).ConclusionsImproved dose metrics with daily adaption could translate to reduced toxicity while maintaining tumor control. Training physicists to perform contouring edits could minimize the time physicians are required at adaptive sessions improving clinical efficiency. All emulated adaptive sessions were completed within 30 min however extra time will be required for patient setup, image acquisition, and treatment delivery.

Fully automated segmentation of clinical target volume in cervical cancer from magnetic resonance imaging with convolutional neural network

AbstractPurposeContouring clinical target volume (CTV) from medical images is an essential step for radiotherapy (RT) planning. Magnetic resonance imaging (MRI) is used as a standard imaging modality for CTV segmentation in cervical cancer due to its superior soft‐tissue contrast. However, the delineation of CTV is challenging as CTV contains microscopic extensions that are not clearly visible even in MR images, resulting in significant contour variability among radiation oncologists depending on their knowledge and experience. In this study, we propose a fully automated deep learning–based method to segment CTV from MR images.MethodsOur method begins with the bladder segmentation, from which the CTV position is estimated in the axial view. The superior–inferior CTV span is then detected using an Attention U‐Net. A CTV‐specific region of interest (ROI) is determined, and three‐dimensional (3‐D) blocks are extracted from the ROI volume. Finally, a CTV segmentation map is computed using a 3‐D U‐Net from the extracted 3‐D blocks.ResultsWe developed and evaluated our method using 213 MRI scans obtained from 125 patients (183 for training, 30 for test). Our method achieved (mean ± SD) Dice similarity coefficient of 0.85 ± 0.03 and the 95th percentile Hausdorff distance of 3.70 ± 0.35 mm on test cases, outperforming other state‐of‐the‐art methods significantly (p‐value < 0.05). Our method also produces an uncertainty map along with the CTV segmentation by employing the Monte Carlo dropout technique to draw physician's attention to the regions with high uncertainty, where careful review and manual correction may be needed.ConclusionsExperimental results show that the developed method is accurate, fast, and reproducible for contouring CTV from MRI, demonstrating its potential to assist radiation oncologists in alleviating the burden of tedious contouring for RT planning in cervical cancer.

Automatic treatment planning for cervical cancer radiation therapy using direct three‐dimensional patient anatomy match

AbstractPurposeCurrent knowledge‐based planning methods for radiation therapy mainly use low‐dimensional features extracted from contoured structures to identify geometrically similar patients. Here, we propose a knowledge‐based treatment planning method where the anatomical similarity is quantified by the rigid registration of the three‐dimensional (3D) planning target volume (PTV) and organs at risks (OARs) between an incoming patient and database patients.MethodsA database that contains PTV and OARs contours from 81 cervical cancer radiation therapy patients was established. To identify the anatomically similar patients, the PTV of the new patient was registered to each PTV in the database and the Dice similarity coefficients were calculated for the PTV, rectum, and bladder between the new patient and database patients. Then the top 20 patients in the PTV match and top 3 patients in the subsequent bladder or rectum match were selected. The best dose–volume histogram parameters from the top three patients were applied as the dose constraints to the automatic plan optimization. A fast Fourier transform algorithm was developed to accelerate the 3D PTV registration process run through the database. The entire treatment planning process was automated using in‐house customized Pinnacle scripts. The automatic plans were generated for 20 patients using leave‐one‐out scheme and were evaluated against the corresponding clinical plans.ResultsThe automatic plans significantly reduced rectum and bladder by 11.79% ± 5.2% (p < 0.01) and 2.85% ± 3.16% (p < 0.01), respectively. The dose parameters achieved for the PTV and other OARs were comparable to those in the clinical plans. The entire planning process, including both dose prediction and inverse optimization, costs about 6 min.ConclusionsThe direct 3D contour match method utilizes the full spatial information of the PTV and OARs of interest and provides an intuitive measurement for patient plan anatomy similarity. The proposed automatic planning method can generate plans with better quality and higher efficiency.

RefineNet‐based 2D and 3D automatic segmentations for clinical target volume and organs at risks for patients with cervical cancer in postoperative radiotherapy

AbstractPurposeAn accurate and reliable target volume delineation is critical for the safe and successful radiotherapy. The purpose of this study is to develop new 2D and 3D automatic segmentation models based on RefineNet for clinical target volume (CTV) and organs at risk (OARs) for postoperative cervical cancer based on computed tomography (CT) images.MethodsA 2D RefineNet and 3D RefineNetPlus3D were adapted and built to automatically segment CTVs and OARs on a total of 44 222 CT slices of 313 patients with stage I–III cervical cancer. Fully convolutional networks (FCNs), U‐Net, context encoder network (CE‐Net), UNet3D, and ResUNet3D were also trained and tested with randomly divided training and validation sets, respectively. The performances of these automatic segmentation models were evaluated by Dice similarity coefficient (DSC), Jaccard similarity coefficient, and average symmetric surface distance when comparing them with manual segmentations with the test data.ResultsThe DSC for RefineNet, FCN, U‐Net, CE‐Net, UNet3D, ResUNet3D, and RefineNet3D were 0.82, 0.80, 0.82, 0.81, 0.80, 0.81, and 0.82 with a mean contouring time of 3.2, 3.4, 8.2, 3.9, 9.8, 11.4, and 6.4 s, respectively. The generated RefineNetPlus3D demonstrated a good performance in the automatic segmentation of bladder, small intestine, rectum, right and left femoral heads with a DSC of 0.97, 0.95, 091, 0.98, and 0.98, respectively, with a mean computation time of 6.6 s.ConclusionsThe newly adapted RefineNet and developed RefineNetPlus3D were promising automatic segmentation models with accurate and clinically acceptable CTV and OARs for cervical cancer patients in postoperative radiotherapy.

A feasibility study of a modified treatment strategy combined external beam radiation therapy and brachytherapy for cervical cancer

Abstract Purpose To evaluate the feasibility of a modified treatment strategy combined external beam radiation therapy (EBRT) and brachytherapy (BT) for cervical cancer through a dosimetry analysis. Material and methods This study retrospectively selected 12 cervical cancer patients treated with the conventional treatment strategy, which consisted of 45─50 Gy/25 fractions of EBRT using volumetric‐modulated arc therapy (VMAT) and image‐guided BT with a fraction dose of 5─7 Gy. The modified treatment strategy decreased the central EBRT dose while increasing the number of BT fractions. New target volumes were additionally contoured, and new VMAT EBRT plans were generated for the modified treatment strategy. The dosimetric parameters for evaluation included the doses to the most irradiated 2 cc (D2cc) of the organs at risk (OARs) and doses to at least 90% (D90) of the gross tumor volume (GTV) and high‐risk clinical target volume (HR‐CTV). The total doses to OARs and targets obtained by adding the equivalent doses in 2 Gy fraction (EQD2) from the EBRT and BT plans were used for quantitative comparison between the modified and conventional treatment strategies. Results Comparison to the conventional treatment strategy, the modified treatment strategy resulted in a higher bladder D2cc, a slightly lower rectal D2cc and a similar HR‐CTV D90, all with no significant differences ( p  > 0.05). The GTV D90 of the modified treatment strategy was significantly higher than that of the conventional treatment strategy ( p  < 0.01). Conclusion The modified treatment strategy can significantly increase the BT dose while remaining the total doses to the bladder and rectum basically unchanged, demonstrating its feasibility and promising prospect in clinical use.

Improvement of treatment plan quality with modified fixed field volumetric modulated arc therapy in cervical cancer

AbstractPurposeThis study aims to introduce modified fixed field volumetric modulated arc therapy (MF‐VMAT) which manually opened the field size by fixing the jaws and comparing it to the typical planning technique, auto field volumetric modulated arc therapy (AF‐VMAT) in cervical cancer treatment planning.Methods and materialsPreviously treated twenty‐eight cervical cancer plans were retrospectively randomly selected and replanned in this study using two different planning techniques: AF‐VMAT and MF‐VMAT, resulting in a total of fifty‐six treatment plans. In this study, we compared both planning techniques in three parts: (1) Organ at Risk (OARs) and whole‐body dose, (2) Treatment plan efficiency, and (3) Treatment plan accuracy.ResultsFor OARs dose, bowel bag (p‐value = 0.001), rectum (p‐value = 0.002), and left femoral head (p‐value = 0.001) and whole‐body (p‐value = 0.000) received a statistically significant dose reduction when using the MF‐VMAT plan. Regarding plan efficiency, MF‐VMAT exhibited a statistically significant increase in both number of monitor units (MUs) and control points (p‐values = 0.000), while beam‐on time, maximum leaf travel, average maximum leaf travel, and maximum leaf travel per gantry rotation were statistically significant decreased (p‐values = 0.000). In terms of plan accuracy, the average gamma passing rate was higher in the MF‐VMAT plan for both absolute dose (AD) (p‐value = 0.001, 0.004) and relative dose (RD) (p‐value = 0.000, 0.000) for 3%/3 and 3%/2 mm gamma criteria, respectively.ConclusionThe MF‐VMAT planning technique significantly reduces OAR doses and decreases the spread of low doses to normal tissues in cervical cancer patients. Additionally, this planning approach demonstrates efficient plans with lower beam‐on time and reduced maximum leaf travel. Furthermore, it indicates higher plan accuracy through an increase in the average gamma passing rate compared to the AF‐VMAT plan. Consequently, MF‐VMAT offers an effective treatment planning technique for cervical cancer patients.

Evaluating automatically generated normal tissue contours for safe use in head and neck and cervical cancer treatment planning

AbstractPurposeVolumetric‐modulated arc therapy (VMAT) is a widely accepted treatment method for head and neck (HN) and cervical cancers; however, creating contours and plan optimization for VMAT plans is a time‐consuming process. Our group has created an automated treatment planning tool, the Radiation Planning Assistant (RPA), that uses deep learning models to generate organs at risk (OARs), planning structures and automates plan optimization. This study quantitatively evaluates the quality of contours generated by the RPA tool.MethodsFor patients with HN (54) and cervical (39) cancers, we retrospectively generated autoplans using the RPA. Autoplans were generated using deep‐learning and RapidPlan models developed in‐house. The autoplans were, then, applied to the original, physician‐drawn contours, which were used as a ground truth (GT) to compare with the autocontours (RPA). Using a “two one‐sided tests” (TOST) procedure, we evaluated whether the autocontour normal tissue dose was equivalent to that of the ground truth by a margin, δ, that we determined based on clinical judgement. We also calculated the number of plans that met established clinically accepted dosimetric criteria.ResultsFor HN plans, 91.8% and 91.7% of structures met dosimetric criteria for automatic and manual contours, respectively; for cervical plans, 95.6% and 95.7% of structures met dosimetric criteria for automatic and manual contours, respectively. Autocontours were equivalent to the ground truth for 71% and 75% of common DVH metrics for the HN and cervix, respectively.ConclusionsThis study shows that dosimetrically equivalent normal tissue contours can be created for HN and cervical cancers using deep learning techniques. In general, differences between the contours did not affect the passing or failing of clinical dose tolerances.

Evaluation of deep learning‐based auto‐segmentation algorithms for delineating clinical target volume and organs at risk involving data for 125 cervical cancer patients

AbstractObjectiveTo evaluate the accuracy of a deep learning‐based auto‐segmentation mode to that of manual contouring by one medical resident, where both entities tried to mimic the delineation "habits" of the same clinical senior physician.MethodsThis study included 125 cervical cancer patients whose clinical target volumes (CTVs) and organs at risk (OARs) were delineated by the same senior physician. Of these 125 cases, 100 were used for model training and the remaining 25 for model testing. In addition, the medical resident instructed by the senior physician for approximately 8 months delineated the CTVs and OARs for the testing cases. The dice similarity coefficient (DSC) and the Hausdorff Distance (HD) were used to evaluate the delineation accuracy for CTV, bladder, rectum, small intestine, femoral‐head‐left, and femoral‐head‐right.ResultsThe DSC values of the auto‐segmentation model and manual contouring by the resident were, respectively, 0.86 and 0.83 for the CTV (P < 0.05), 0.91 and 0.91 for the bladder (P > 0.05), 0.88 and 0.84 for the femoral‐head‐right (P < 0.05), 0.88 and 0.84 for the femoral‐head‐left (P < 0.05), 0.86 and 0.81 for the small intestine (P < 0.05), and 0.81 and 0.84 for the rectum (P > 0.05). The HD (mm) values were, respectively, 14.84 and 18.37 for the CTV (P < 0.05), 7.82 and 7.63 for the bladder (P > 0.05), 6.18 and 6.75 for the femoral‐head‐right (P > 0.05), 6.17 and 6.31 for the femoral‐head‐left (P > 0.05), 22.21 and 26.70 for the small intestine (P > 0.05), and 7.04 and 6.13 for the rectum (P > 0.05). The auto‐segmentation model took approximately 2 min to delineate the CTV and OARs while the resident took approximately 90 min to complete the same task.ConclusionThe auto‐segmentation model was as accurate as the medical resident but with much better efficiency in this study. Furthermore, the auto‐segmentation approach offers additional perceivable advantages of being consistent and ever improving when compared with manual approaches.

Risk for second bladder and rectal malignancies from cervical cancer irradiation

AbstractThe objective of this study was to estimate the risk of developing second malignancies to partially in‐field organs from volumetric modulated arc therapy (VMAT) of cervical cancer and to compare the above risks with those from the conventional three‐dimensional conformal radiotherapy (3D‐CRT). Seventeen consecutive patients with uterine cervix carcinoma were selected. VMAT and 3D‐CRT plans were generated with 6 and 10 MV photons, respectively. The prescribed tumor dose was 45 Gy given in 25 fractions. Differential dose‐volume histogram data from the treatment plans were obtained for the partially in‐field organs such as bladder and rectum. These data were used to estimate the patient‐specific lifetime attributable risk (LAR) for bladder and rectal cancer induction with a non‐linear model based on a mixture of plateau and bell‐shaped dose–response relationships. The estimated risks per 10000 people were compared with the baseline risks for unexposed population. The patient‐specific rectal cancer risk estimates from VMAT were significantly lower than those from 3D‐CRT (P = 0.0144). The LARs for developing bladder malignancies from VMAT were significantly high compared to those from conventional irradiation (P = 0.0003). The mean difference between the patient‐specific LARs for radiation‐induced bladder and rectal malignancies as derived from 3D‐CRT and VMAT plans was 6.6% and 2.0%, respectively. The average LAR for developing bladder and rectal malignant diseases due to VMAT was 9.2 × 10‐4 and 43.7 × 10‐4, respectively. The corresponding risks following 3D‐CRT were 8.6 × 10‐4 and 44.6 × 10‐4. These average risks showed that pelvic irradiation increases the baseline probability for cancer induction by 12.6‐19.1%. The differences in the second cancer risks associated with the VMAT and 3D‐CRT for cervical cancer were found to be small. Both treatment techniques resulted in considerable increased probabilities for developing bladder and rectal malignancies relative to those of unirradiated population.

Deep learning‐based auto‐segmentation of clinical target volumes for radiotherapy treatment of cervical cancer

AbstractObjectivesBecause radiotherapy is indispensible for treating cervical cancer, it is critical to accurately and efficiently delineate the radiation targets. We evaluated a deep learning (DL)‐based auto‐segmentation algorithm for automatic contouring of clinical target volumes (CTVs) in cervical cancers.MethodsComputed tomography (CT) datasets from 535 cervical cancers treated with definitive or postoperative radiotherapy were collected. A DL tool based on VB‐Net was developed to delineate CTVs of the pelvic lymph drainage area (dCTV1) and parametrial area (dCTV2) in the definitive radiotherapy group. The training/validation/test number is 157/20/23. CTV of the pelvic lymph drainage area (pCTV1) was delineated in the postoperative radiotherapy group. The training/validation/test number is 272/30/33. Dice similarity coefficient (DSC), mean surface distance (MSD), and Hausdorff distance (HD) were used to evaluate the contouring accuracy. Contouring times were recorded for efficiency comparison.ResultsThe mean DSC, MSD, and HD values for our DL‐based tool were 0.88/1.32 mm/21.60 mm for dCTV1, 0.70/2.42 mm/22.44 mm for dCTV2, and 0.86/1.15 mm/20.78 mm for pCTV1. Only minor modifications were needed for 63.5% of auto‐segmentations to meet the clinical requirements. The contouring accuracy of the DL‐based tool was comparable to that of senior radiation oncologists and was superior to that of junior/intermediate radiation oncologists. Additionally, DL assistance improved the performance of junior radiation oncologists for dCTV2 and pCTV1 contouring (mean DSC increases: 0.20 for dCTV2, 0.03 for pCTV1; mean contouring time decrease: 9.8 min for dCTV2, 28.9 min for pCTV1).ConclusionsDL‐based auto‐segmentation improves CTV contouring accuracy, reduces contouring time, and improves clinical efficiency for treating cervical cancer.

Initial analysis of the dosimetric benefit and clinical resource cost of CBCT‐based online adaptive radiotherapy for patients with cancers of the cervix or rectum

AbstractPurposeThis provides a benchmark of dosimetric benefit and clinical cost of cone‐beam CT‐based online adaptive radiotherapy (ART) technology for cervical and rectal cancer patients.MethodsAn emulator of a CBCT‐based online ART system was used to simulate more than 300 treatments for 13 cervical and 15 rectal cancer patients. CBCT images were used to generate adaptive replans. To measure clinical resource cost, the six phases of the workflow were timed. To evaluate the dosimetric benefit, changes in dosimetric values were assessed. These included minimum dose (Dmin) and volume receiving 95% of prescription (V95%) for the planning target volume (PTV) and the clinical target volume (CTV), and maximum 2 cc's (D2cc) of the bladder, bowel, rectum, and sigmoid colon.ResultsThe average duration of the workflow was 24.4 and 9.2 min for cervical and rectal cancer patients, respectively. A large proportion of time was dedicated to editing target contours (13.1 and 2.7 min, respectively). For cervical cancer patients, the replan changed the Dmin to the PTVs and CTVs for each fraction 0.25 and 0.25 Gy, respectively. The replan changed the V95% by 9.2 and 7.9%. The D2cc to the bladder, bowel, rectum, and sigmoid colon for each fraction changed −0.02, −0.08, −0.07, and −0.04 Gy, respectively. For rectal cancer patients, the replan changed the Dmin to the PTVs and CTVs for each fraction of 0.20 and 0.24 Gy, respectively. The replan changed the V95% by 4.1 and 1.5%. The D2cc to the bladder and bowel for each fraction changed 0.02 and −0.02 Gy, respectively.ConclusionsDosimetric benefits can be achieved with CBCT‐based online ART that is amenable to conventional appointment slots. The clinical significance of these benefits remains to be determined. Managing contours was the primary factor affecting the total duration and is imperative for safe and effective adaptive radiotherapy.

Three‐dimensional deep neural network for automatic delineation of cervical cancer in planning computed tomography images

AbstractPurposeRadiation therapy is an essential treatment modality for cervical cancer, while accurate and efficient segmentation methods are needed to improve the workflow. In this study, a three‐dimensional V‐net model is proposed to automatically segment clinical target volume (CTV) and organs at risk (OARs), and to provide prospective guidance for low lose area.Material and methodsA total of 130 CT datasets were included. Ninety cases were randomly selected as the training data, with 10 cases used as the validation data, and the remaining 30 cases as testing data. The V‐net model was implemented with Tensorflow package to segment the CTV and OARs, as well as regions of 5 Gy, 10 Gy, 15 Gy, and 20 Gy isodose lines covered. The auto‐segmentation by V‐net was compared to auto‐segmentation by U‐net. Four representative parameters were calculated to evaluate the accuracy of the delineation, including Dice similarity coefficients (DSC), Jaccard index (JI), average surface distance (ASD), and Hausdorff distance (HD).ResultsThe V‐net and U‐net achieved the average DSC value for CTV of 0.85 and 0.83, average JI values of 0.77 and 0.75, average ASD values of 2.58 and 2.26, average HD of 11.2 and 10.08, respectively. As for the OARs, the performance of the V‐net model in the colon was significantly better than the U‐net model (p = 0.046), and the performance in the kidney, bladder, femoral head, and pelvic bones were comparable to the U‐net model. For prediction of low‐dose areas, the average DSC of the patients’ 5 Gy dose area in the test set were 0.88 and 0.83, for V‐net and U‐net, respectively.ConclusionsIt is feasible to use the V‐Net model to automatically segment cervical cancer CTV and OARs to achieve a more efficient radiotherapy workflow. In the delineation of most target areas and OARs, the performance of V‐net is better than U‐net. It also offers advantages with its feature of predicting the low‐dose area prospectively before radiation therapy (RT).