Physics in Medicine & Biology

Deciphering oxygen distribution and hypoxia profiles in the tumor microenvironment: a data-driven mechanistic modeling approach

Abstract Objective . The distribution of hypoxia within tissues plays a critical role in tumor diagnosis and prognosis. Recognizing the significance of tumor oxygenation and hypoxia gradients, we introduce mathematical frameworks grounded in mechanistic modeling approaches for their quantitative assessment within a tumor microenvironment. By utilizing known blood vasculature, we aim to predict hypoxia levels across different tumor types. Approach . Our approach offers a computational method to measure and predict hypoxia using known blood vasculature. By formulating a reaction-diffusion model for oxygen distribution, we derive the corresponding hypoxia profile. Main results . The framework successfully replicates observed inter- and intra-tumor heterogeneity in experimentally obtained hypoxia profiles across various tumor types (breast, ovarian, pancreatic). Additionally, we propose a data-driven method to deduce partial differential equation models with spatially dependent parameters, which allows us to comprehend the variability of hypoxia profiles within tissues. The versatility of our framework lies in capturing diverse and dynamic behaviors of tumor oxygenation, as well as categorizing states of vascularization based on the dynamics of oxygen molecules, as identified by the model parameters. Significance . The proposed data-informed mechanistic method quantitatively assesses hypoxia in the tumor microenvironment by integrating diverse histopathological data and making predictions across different types of data. The framework provides valuable insights from both modeling and biological perspectives, advancing our comprehension of spatio-temporal dynamics of tumor oxygenation.

Contrast-enhanced photon-counting micro-CT of tumor xenograft models

Abstract Objective . Photon-counting micro-computed tomography (micro-CT) is a major advance in small animal preclinical imaging. Small molecule- and nanoparticle-based contrast agents have been widely used to enable the differentiation of liver tumors from surrounding tissues using photon-counting micro-CT. However, there is a notable gap in the application of these market-available agents to the imaging of breast and ovarian tumors using photon-counting micro-CT. Herein, we have used photon-counting micro-CT to determine the effectiveness of these contrast agents in differentiating ovarian and breast tumor xenografts in live, intact mice. Approach . Nude mice carrying different types of breast and ovarian tumor xenografts (AU565, MDA-MB-231 and SKOV-3 human cancer cells) were injected with ISOVUE-370 (a small molecule-based agent) or Exitron Nano 12000 (a nanoparticle-based agent) and subjected to photon-counting micro-CT. To improve tumor visualization using photon-counting micro-CT, we developed a novel color visualization method, which changes color tones to highlight contrast media distribution, offering a robust alternative to traditional material decomposition methods with less computational demand. Main results . Our in vivo experiments confirm the effectiveness of this color visualization approach, showing distinct enhancement characteristics for each contrast agent. Qualitative and quantitative analyses suggest that Exitron Nano 12000 provides superior vasculature enhancement and better quantitative consistency across scans, while ISOVUE-370 delivers a more comprehensive tumor enhancement but with significant variance between scans due to its short blood half-time. Further, a paired t-test on mean and standard deviation values within tumor volumes showed significant differences between the AU565 and SKOV-3 tumor models with the nanoparticle-based contrast agent ( p -values < 0.02), attributable to their distinct vascularity, as confirmed by immunohistochemical analysis. Significance . These findings underscore the utility of photon-counting micro-CT in non-invasively assessing the morphology and anatomy of different tumor xenografts, which is crucial for tumor characterization and longitudinal monitoring of tumor progression and response to treatments.

Predicting treatment plan approval probability for high-dose-rate brachytherapy of cervical cancer using adversarial deep learning

Abstract Objective. Predicting the probability of having the plan approved by the physician is important for automatic treatment planning. Driven by the mathematical foundation of deep learning that can use a deep neural network to represent functions accurately and flexibly, we developed a deep-learning framework that learns the probability of plan approval for cervical cancer high-dose-rate brachytherapy (HDRBT). Approach. The system consisted of a dose prediction network (DPN) and a plan-approval probability network (PPN). DPN predicts organs at risk (OAR) D 2cc and CTV D 90% of the current fraction from the patient’s current anatomy and prescription dose of HDRBT. PPN outputs the probability of a given plan being acceptable to the physician based on the patients anatomy and the total dose combining HDRBT and external beam radiotherapy sessions. Training of the networks was achieved by first training them separately for a good initialization, and then jointly via an adversarial process. We collected approved treatment plans of 248 treatment fractions from 63 patients. Among them, 216 plans from 54 patients were employed in a four-fold cross validation study, and the remaining 32 plans from other 9 patients were saved for independent testing. Main results. DPN predicted equivalent dose of 2 Gy for bladder, rectum, sigmoid D 2cc and CTV D 90% with a relative error of 11.51% ± 6.92%, 8.23% ± 5.75%, 7.12% ± 6.00%, and 10.16% ± 10.42%, respectively. In a task that differentiates clinically approved plans and disapproved plans generated by perturbing doses in ground truth approved plans by 20%, PPN achieved accuracy, sensitivity, specificity, and area under the curve 0.70, 0.74, 0.65, and 0.74. Significance. We demonstrated the feasibility of developing a novel deep-learning framework that predicts a probability of plan approval for HDRBT of cervical cancer, which is an essential component in automatic treatment planning.

Toward fully automated pre-implantation planning for cervical cancer brachytherapy: a template-guided multi-criteria optimization framework for catheter placement and dose distribution

Abstract Background and purpose. Image-guided intracavitary/interstitial brachytherapy is standard for locally advanced cervical cancer (LACC), yet the lack of pre-implantation planning often leads to suboptimal catheter placement and dose distribution. A template-guided, multi-criteria optimization-based pre-implantation planning system was thus proposed, which integrates dosimetric, radiobiological, and geometric objectives. Materials and methods. The developed system employs an improved wish-list optimization strategy featuring a tighten-relax mechanism with hybrid constraints to enhance robustness and resolve conflicts from impractical wish-lists. It incorporates dose-volume indices, the generalized equivalent uniform dose, and a novel total conformity index, while supporting oblique catheter insertions for improved dose conformation. Dosimetric performance was benchmarked against clinical approved plans in 40 LACC cases, focusing on target coverage, conformity, and organ-at-risk sparing. Results. Compared to the catheter configurations in clinically approved plans, the proposed method resulted in catheter displacement exceeding 5 mm in 11 out of 40 LACC cases and catheter angular deviation greater than 10°in 7 cases. Correspondingly, the proportion of new plans meeting the D 90 > 100% criterion was 100% with template guidance versus 92.5% without, both significantly higher than the 82.5% achieved by original clinical plans. The method also increased V 90 and V 100 by 3.4% and 4.9% (both p < 0.001), improved dose conformity from 0.59 to 0.61 ( p < 0.001), while maintaining OAR D 2cc within clinical limits. Conclusion. The proposed pre-implantation planning method reduces reliance on operator experience and offers a robust, automated solution for high-conformity brachytherapy.

Predicting lymphocyte dose and surviving fraction for VMAT and IMPT treatments with a dynamic lymphocyte flow model for locally advanced cervical cancer

Abstract Objective. A dynamic model is developed to predict the impact of radiotherapy on circulating lymphocyte counts in women with locally advanced cervical cancer (LACC). This study aims to compare the effects of photon and proton therapy, as well as the influence of bone marrow sparing (BMS) techniques, on relative lymphocyte preservation over time. Approach. A dynamic lymphocyte flow model was developed to simulate the migration of lymphocytes based on seven compartments. Biological cell death and lymphocyte production were integrated across compartments. The lymphocyte flow model was applied to 19 LACC patients. Volumetric modulated arc therapy (VMAT) and intensity modulated proton therapy (IMPT) treatment plans were created for each patient without BMS and with BMS. The model calculated radiation dose to lymphocytes to estimate radiation-induced cell death over time. The output of the model was the relative lymphocyte count relative to baseline (RLC) over time and the RLC nadir in the blood and total body. Main results. According to the model, IMPT resulted in lower doses to lymphocyte and higher RLC nadirs compared to VMAT for all 19 patients. The total RLC nadir (mean ± SD) was 48.4% ± 4.0% for VMAT and 62.5% ± 5.1% for IMPT. In the blood compartment, the RLC nadir was 32.7% ± 3.5% for VMAT and 47.7% ± 5.9% for IMPT. The RLC nadir in the blood compartment improved with 3 Gy BMS from 32.7% ± 3.5% to 33.0% ± 3.5% , while it decreased for IMPT from 47.7% ± 5.9% to 46.6% ± 6.0%. Total RLC nadir decreased with BMS for VMAT from 48.4% ± 4.0% to 48.2% ± 3.9% and for IMPT from 62.5% ± 5.1% to 60.9% ± 5.3%. Significance. By incorporating a dynamic flow model, we predicted the RLC over time. The model predicted a substantial sparing effect IMPT has on the lymphocytes compared to VMAT. This sparing was both present in the blood and the total body. Sparing the bone marrow showed only a minimal effect on the RLC.

Quantitative and automatic plan-of-the-day assessment to facilitate adaptive radiotherapy in cervical cancer

Abstract Objective. To facilitate implementation of plan-of-the-day (POTD) selection for treating locally advanced cervical cancer (LACC), we developed a POTD assessment tool for CBCT-guided radiotherapy (RT). A female pelvis segmentation model (U-Seg3) is combined with a novel quantitative standard operating procedure (qSOP) to identify optimal and acceptable plans. Approach. The planning CT[i], corresponding structure set[ii], and manually contoured CBCTs[iii] (n = 226) from 39 LACC patients treated with POTD (n = 11) or non-adaptive RT (n = 28) were used to develop U-Seg3, an algorithm incorporating deep-learning and deformable image registration techniques to segment the low-risk clinical target volume (LR-CTV), high-risk CTV (HR-CTV), bladder, rectum, and bowel bag. A single-channel input model (iii only, U-Seg1) was also developed. Contoured CBCTs from the POTD patients were (a) reserved for U-Seg3 validation/testing, (b) audited to determine optimal and acceptable plans, and (c) used to empirically derive a qSOP that maximised classification accuracy. Main results. The median (interquartile range) dice similarity coefficient (DSC) between manual and U-Seg3 contours was 0.83 [0.80], 0.78 [0.13], 0.94 [0.05], 0.86 [0.09], and 0.90 [0.05] for the LR-CTV, HR-CTV, bladder, rectum, and bowel. These were significantly higher than U-Seg1 in all structures but bladder. The qSOP classified plans as acceptable if they met target coverage thresholds (LR-CTV ⩾ 99%, HR-CTV ⩾ 99.8%), with lower LR-CTV coverage ( ⩾ 95%) sometimes allowed. The acceptable plan minimizing bowel irradiation was considered optimal unless substantial bladder sparing could be achieved. With U-Seg3 embedded in the qSOP, optimal and acceptable plans were identified in 46/60 and 57/60 cases. Significance. U-Seg3 outperforms U-Seg1 and all known CBCT-based segmentation models of the female pelvis both in terms of scope and accuracy (median DSC improvement ranging from 0.03–0.06). The tool combining U-Seg3 and the qSOP identifies optimal plans with equivalent accuracy as two observers. In an implementation strategy whereby this tool serves as the second observer, plan selection confidence and decision-making time could be improved whilst simultaneously reducing the required number of POTD-trained radiographers by 50%.

Margin and robustness settings for a library-of-plans IMPT strategy for locally advanced cervical cancer

Abstract Objective. This study aims to determine a margin and robustness setting for treating locally advanced cervical cancer (LACC) with a library-of-plans (LoP) based online-adaptive intensity-modulated proton therapy (IMPT). Approach. We analyzed 13 LACC patients with delineated planning and weekly repeat CT scans (reCTs). For each patient, 120 IMPT treatments of 25 fractions were simulated with a LoPs approach. Six different robustness settings (2–7 mm set-up robustness (SR) plus 3% range robustness (RR)) were used to create those 120 IMPT plans. Each fraction was simulated with a weekly reCT, combined with the sampling of inter- and intrafraction treatment uncertainties. The fraction doses were accumulated to obtain a treatment dose to the target volumes, distinguishing between the low-risk clinical target volume (CTV-T-LR) and the elective CTV (CTV-E). If one of the two targets obtained an adequate coverage for more than 90% of the treatments, different anisotropic margins were sampled on top of the robustness setting to the other target to obtain the Pareto-optimal margin in terms of adequate coverage versus increase in target volume. Main results. The percentage of treatments that reach the dose criterion V 42.75Gy > 95% for the CTV-T-LR was 22.3%, 28.5%, 51.2%, 73.1%, 85.3%, and 90.0% for 2, 3, 4, 5, 6, and 7 mm SR plus 3% RR and for the CTV-E, this percentage was 60.4%, 73.8%, 86.5%, 92.3%, 96.9%, and 98.5%. The Pareto-optimal margin combined with a 5 mm/3% robustness setting for the CTV-T-LR with an adequate coverage for >90% of the treatments was given by {0, 1, 0, 3, 3, 0} mm in the left, right, anterior, posterior, cranial, caudal direction. Significance. Our study evaluated combinations of robustness and anisotropic margin settings for IMPT for LACC. With 5 mm SR and 3% RR for CTV-E and CTV-T-LR plus a margin to the CTV-T-LR of {0, 1, 0, 3, 3, 0} mm in left, right, anterior, posterior, cranial, and caudal ensured an adequate coverage for >90% of the simulated IMPT treatments.

Automated planning of curved needle channels in 3D printed patient-tailored applicators for cervical cancer brachytherapy

Abstract Purpose. Patient-tailored intracavitary/interstitial (IC/IS) brachytherapy (BT) applicators may increase dose conformity in cervical cancer patients. Current configuration planning methods in these custom applicators rely on manual specification or a small set of (straight) needles. This work introduces and validates a two-stage approach for establishing channel configurations in the 3D printed patient-tailored ARCHITECT applicator. Methods. For each patient, the patient-tailored applicator shape was based on the first BT application with a commercial applicator and integrated connectors to a commercial (Geneva) intrauterine tube and two lunar ring channels. First, a large candidate set was generated of channels that steer the needle to desired poses in the target region and are contained in the applicator. The channels’ centrelines were represented by Bézier curves. Channels running between straight target segments and entry points were optimised and refined to ensure (dynamic) feasibility. Second, channel configurations were selected using geometric coverage optimisation. This workflow was applied to establish patient-tailored geometries for twenty-two patients previously treated using the Venezia applicator. Treatment plans were automatically generated using the in-house developed algorithm BiCycle. Plans for the clinically used configuration, T P clin , and patient-tailored configuration, T P arch , were compared. Results. Channel configurations could be generated in clinically feasible time (median: 2651 s, range 1826–3812 s). All T P arch and T P clin plans were acceptable, but planning aims were more frequently attained with patient-tailored configurations (115/132 versus 100/132 instances). Median CTVIR D 98 and bladder D 2 c m 3 doses significantly improved ( p < 0.001 and p < 0.01 respectively) in T P arch plans in comparison with T P clin plans, and in approximately half of the patients dosimetric indices improved. Conclusion. Automated patient-tailored BT channel configuration planning for 3D printed applicators is clinically feasible. A treatment planning study showed that all plans met planning limits for the patient-tailored configurations, and in selected cases improved the plan quality in comparison with commercial applicator configurations.

Relative biological effectiveness of clinically relevant photon energies for the survival of human colorectal, cervical, and prostate cancer cell lines

Abstract Objective. Relative biological effectiveness (RBE) differs between radiation qualities. However, an RBE of 1.0 has been established for photons regardless of the wide range of photon energies used clinically, the lack of reproducibility in radiobiological studies, and outdated reference energies used in the experimental literature. Moreover, due to intrinsic radiosensitivity, different cancer types have different responses to radiation. This study aimed to characterize the RBE of clinically relevant high and low photon energies in vitro for three human cancer cell lines: HCT116 (colon), HeLa (cervix), and PC3 (prostate). Approach. Experiments were conducted following dosimetry protocols provided by the American Association of Physicists in Medicine. Cells were irradiated with 6 MV x-rays, an 192Ir brachytherapy source, 225 kVp and 50 kVp x-rays. Cell survival post-irradiation was assessed using the clonogenic assay. Survival fractions were fitted using the linear quadratic model, and survival curves were generated for RBE calculations. Main results. Cell killing was more efficient with decreasing photon energy. Using 225 kVp x-rays as the reference, the HCT116 RBESF0.1 for 6 MV x-rays, 192Ir, and 50 kVp x-rays were 0.89 ± 0.03, 0.95 ± 0.03, and 1.24 ± 0.04; the HeLa RBESF0.1 were 0.95 ± 0.04, 0.97 ± 0.05, and 1.09 ± 0.03, and the PC3 RBESF0.1 were 0.84 ± 0.01, 0.84 ± 0.01, and 1.13 ± 0.02, respectively. HeLa and PC3 cells had varying radiosensitivity when irradiated with 225 and 50 kVp x-rays. Significance. This difference supports the notion that RBE may not be 1.0 for all photons through experimental investigations that employed precise dosimetry. It highlights that different cancer types may not have identical responses to the same irradiation quality. Additionally, the RBE of clinically relevant photons was updated to the reference energy of 225 kVp x-rays.

DMGM: deformable-mechanism based cervical cancer staging via MRI multi-sequence *

Abstract Objective. This study aims to leverage a deep learning approach, specifically a deformable convolutional layer, for staging cervical cancer using multi-sequence MRI images. This is in response to the challenges doctors face in simultaneously identifying multiple sequences, a task that computer-aided diagnosis systems can potentially improve due to their vast information storage capabilities. Approach. To address the challenge of limited sample sizes, we introduce a sequence enhancement strategy to diversify samples and mitigate overfitting. We propose a novel deformable ConvLSTM module that integrates a deformable mechanism with ConvLSTM, enabling the model to adapt to data with varying structures. Furthermore, we introduce the deformable multi-sequence guidance model (DMGM) as an auxiliary diagnostic tool for cervical cancer staging. Main results. Through extensive testing, including comparative and ablation studies, we validate the effectiveness of the deformable ConvLSTM module and the DMGM. Our findings highlight the model’s ability to adapt to the deformation mechanism and address the challenges in cervical cancer tumor staging, thereby overcoming the overfitting issue and ensuring the synchronization of asynchronous scan sequences. The research also utilized the multi-modal data from BraTS 2019 as an external test dataset to validate the effectiveness of the proposed methodology presented in this study. Significance. The DMGM represents the first deep learning model to analyze multiple MRI sequences for cervical cancer, demonstrating strong generalization capabilities and effective staging in small dataset scenarios. This has significant implications for both deep learning applications and medical diagnostics. The source code will be made available subsequently.

Photodynamic therapy in 2D and 3D human cervical carcinoma cell cultures employing LED light sources emitting at different wavelengths

Abstract Light of different wavelengths can be used to obtain a more profitable outcome of photodynamic therapy (PDT), according to the absorption bands of the photosensitizer (PS). Low-grade cervical intraepithelial neoplasias (CINs) are superficial lesions that can be treated with light of shorter wavelength than red because a large light penetration depth in tissue is not necessary. We report a comparative investigation performed to evaluate the efficacy of light-emitting diodes (LEDs) of different wavelengths in the photodynamic treatment applied to both 2D and 3D HeLa cell spheroid cultures. The spheroids are utilized as a PDT dosage model, and cell viability is evaluated at different sections of the spheroids by confocal microscopy. Cells incubated with m-tetrahydroxyphenyl chlorin are illuminated with LED systems working in the low fluence range, emitting in the violet (390–415 nm), blue (440–470 nm), red (620–645 nm) and deep red (640–670 nm) regions of the light spectrum at various exposures times ( t I ) comprised between 0.5 and 30 min. PDT experiments performed on both 2D and 3D cell cultures indicate that the PDT treatment outcome is more efficient with violet light followed by red light. Dynamic data from the front displacement velocity of large 2D-quasi-radial colonies generated from cell spheroids adhered to the Petri dish bottom as well as the evolution of the 3D growth give further insight about the effect of PDT at each condition. Results from 3D cultures indicate that the penetration of the violet light is appropriate to kill HeLa cells several layers below, showing cell damage and death not only in the outer rim of the illuminated spheroids, where a PS accumulation exists, but also in the more internal region. Results indicate that violet LED light could be useful to treat CINs involving superficial dysplasia.

Incorporating cross-voxel exchange into the analysis of dynamic contrast-enhanced imaging data: theory, simulations and experimental results

Abstract Predictions of tumour perfusion are key determinants of drug delivery and responsiveness to therapy. Pharmacokinetic models allow for the estimation of perfusion properties of tumour tissues but many assume no dispersion associated with tracer transport away from the capillaries and through the tissue. At the level of a voxel, this translates to assuming no cross-voxel tracer exchange, often leading to the misinterpretation of derived perfusion parameters. Tofts model (TM), a compartmental model widely used in oncology, also makes this assumption. A more realistic description is required to quantify kinetic properties of tracers, such as convection and diffusion. We propose a Cross-Voxel Exchange Model (CVXM) for analysing cross-voxel tracer kinetics. In silico datasets quantifying the roles of convection and diffusion in tracer transport (which TM ignores) were employed to investigate the interpretation of Tofts’ perfusion parameters compared to CVXM. TM returned inaccurate values of K t r a n s and v e where diffusive and convective mechanisms are pronounced (up to 20% and 300% error respectively). A mathematical equation, developed in this work, predicts and gives the correct physiological interpretation of Tofts’ v e . Finally, transport parameters were derived from dynamic contrast enhanced-magnetic resonance imaging of a TS-415 human cervical carcinoma xenograft by using TM and CVXM. The latter deduced lower values of K t r a n s and v e compared to TM (lower by up to 63% and 76% respectively). It also allowed the detection of a diffusive flux (mean diffusivity 155 μ m 2 s −1 ) in the tumour tissue, as well as an increased convective flow at the periphery (mean velocity 2.3 μ m s −1 detected). The results serve as a proof of concept establishing the feasibility of using CVXM for accurately determining transport metrics that characterize the exchange of tracer between voxels. CVXM needs to be investigated further as its parameters can be linked to the tumour microenvironment properties (permeability, pressure…), potentially leading to enhanced personalized treatment planning.

Automatic segmentation for plan-of-the-day selection in CBCT-guided adaptive radiation therapy of cervical cancer

Abstract Objective. Plan-of-the-day (PoD) adaptive radiation therapy (ART) is based on a library of treatment plans, among which, at each treatment fraction, the PoD is selected using daily images. However, this strategy is limited by PoD selection uncertainties. This work aimed to propose and evaluate a workflow to automatically and quantitatively identify the PoD for cervix cancer ART based on daily CBCT images. Approach. The quantification was based on the segmentation of the main structures of interest in the CBCT images (clinical target volume [CTV], rectum, bladder, and bowel bag) using a deep learning model. Then, the PoD was selected from the treatment plan library according to the geometrical coverage of the CTV. For the evaluation, the resulting PoD was compared to the one obtained considering reference CBCT delineations. Main results. In experiments on a database of 23 patients with 272 CBCT images, the proposed method obtained an agreement between the reference PoD and the automatically identified PoD for 91.5% of treatment fractions (99.6% when considering a 5% margin on CTV coverage). Significance. The proposed automatic workflow automatically selected PoD for ART using deep-learning methods. The results showed the ability of the proposed process to identify the optimal PoD in a treatment plan library.

Self-supervised learning for multi-center magnetic resonance imaging harmonization without traveling phantoms

Abstract Objective. With the progress of artificial intelligence (AI) in magnetic resonance imaging (MRI), large-scale multi-center MRI datasets have a great influence on diagnosis accuracy and model performance. However, multi-center images are highly variable due to the variety of scanners or scanning parameters in use, which has a negative effect on the generality of AI-based diagnosis models. To address this problem, we propose a self-supervised harmonization (SSH) method. Approach. Mapping the style of images between centers allows harmonization without traveling phantoms to be formalized as an unpaired image-to-image translation problem between two domains. The mapping is a two-stage transform, consisting of a modified cycle generative adversarial network (cycleGAN) for style transfer and a histogram matching module for structure fidelity. The proposed algorithm is demonstrated using female pelvic MRI images from two 3 T systems and compared with three state-of-the-art methods and one conventional method. In the absence of traveling phantoms, we evaluate harmonization from three perspectives: image fidelity, ability to remove inter-center differences, and influence on the downstream model. Main results. The improved image sharpness and structure fidelity are observed using the proposed harmonization pipeline. It largely decreases the number of features with a significant difference between two systems (from 64 to 45, lower than dualGAN: 57, cycleGAN: 59, ComBat: 64, and CLAHE: 54). In the downstream cervical cancer classification, it yields an area under the receiver operating characteristic curve of 0.894 (higher than dualGAN: 0.828, cycleGAN: 0.812, ComBat: 0.685, and CLAHE: 0.770). Significance. Our SSH method yields superior generality of downstream cervical cancer classification models by significantly decreasing the difference in radiomics features, and it achieves greater image fidelity.

Colposcopic multimodal fusion for the classification of cervical lesions

Abstract Objective: Cervical cancer is one of the two biggest killers of women and early detection of cervical precancerous lesions can effectively improve the survival rate of patients. Manual diagnosis by combining colposcopic images and clinical examination results is the main clinical diagnosis method at present. Developing an intelligent diagnosis algorithm based on artificial intelligence is an inevitable trend to solve the objectification of diagnosis and improve the quality and efficiency of diagnosis. Approach: A colposcopic multimodal fusion convolutional neural network (CMF-CNN) was proposed for the classification of cervical lesions. Mask region convolutional neural network was used to detect the cervical region while the encoding network EfficientNet-B3 was introduced to extract the multimodal image features from the acetic image and iodine image. Finally, Squeeze-and-Excitation, Atrous Spatial Pyramid Pooling, and convolution block were also adopted to encode and fuse the patient’s clinical text information. Main results: The experimental results showed that in 7106 cases of colposcopy, the accuracy, macro F1-score, macro-areas under the curve of the proposed model were 92.70%, 92.74%, 98.56%, respectively. They are superior to the mainstream unimodal image classification models. Significance: CMF-CNN proposed in this paper combines multimodal information, which has high performance in the classification of cervical lesions in colposcopy, so it can provide comprehensive diagnostic aid.

BAIRDA: a novel in vitro setup to quantify radiobiological parameters for cervical cancer brachytherapy dose estimations

AbstractObjective. Brachytherapy (BT) dose prescriptions for locally advanced cervical cancer are made with account for the radiobiological parameters,α/βratio and halftime of repair (T1/2). However, a wide range of parameter values has been reported which can challenge commonly held equivalencies between dose prescriptions. This is the first reported study that aims to develop anin vitroexperimental technique using clinical high-dose-rate (HDR) and pulsed-dose-rate (PDR) Ir-192 brachytherapy afterloaders to quantify these parametersin vitroand to contextualize findings within contemporary practice.Approach. To efficiently quantifyα/βandT1/2,in vitroexperiments more reflective of clinical BT practice than traditional clonogenic survival assays were developed and applied to four squamous cell carcinoma cell lines (CaSki, C-33A, SiHa, and SW756). Radiation was delivered using single acute and fractionated dose treatments with a conventional irradiator and clinical HDR and PDR BT afterloaders. For the latter, a novelbrachytherapyafterloaderin vitroradiationdeliveryapparatus (BAIRDA) was developed.Main Results. Theα/βandT1/2values determined using BAIRDA and the conventional irradiator showed close agreement, validating the novel apparatus and technique. For CaSki, C-33A, SiHa, and SW756, the BAIRDA-measuredα/βratios (5.2 [4.6–5.8], 5.6 [4.5–6.6], 6.3 [4.9–7.7], and 5.3 [4.7–6.0] Gy, respectively) were consistently smaller, while theT1/2(3.3 [2.7–3.9], 2.7 [2.0–3.3], 2.8 (2.4–3.1], and 4.8 [4.1–5.4] hours) larger, than the widely accepted values in clinical practice (α/β= 10 Gy;T1/2 = 1.5 h).Significance.In vitroexperiments using BAIRDA provided evidence for differences between the conventionally selected and experimentally determinedα/βratio andT1/2. Treatment regimens using HDR-BT and PDR-BT, designed to deliver equivalent radiobiological doses based on conventional values, were shown to differ by up to 27 Gy EQD2 – an effect that could impact treatment outcomes in cervical cancer. Furthermore, with BAIRDA, we have developed a novel method for radiobiological research in BT.

Optimizing prompt strategies for SAM: advancing lesion segmentation across diverse medical imaging modalities

Abstract Object. To evaluate various segment anything model (SAM) prompt strategies across four lesion datasets and to subsequently develop a reinforcement learning (RL) agent to optimize SAM prompt placement. Approach. This retrospective study included patients with four independent ovarian, lung, renal, and breast tumor datasets. Manual segmentation and SAM-assisted segmentation were performed for all lesions. An RL model was developed to predict and select SAM points to maximize segmentation performance. Statistical analysis of segmentation was conducted using pairwise t-tests. Main results. Results show that increasing the number of prompt points significantly improves segmentation accuracy, with Dice coefficients rising from 0.272 for a single point to 0.806 for five or more points in ovarian tumors. The prompt location also influenced performance, with surface and union-based prompts outperforming center-based prompts, achieving mean Dice coefficients of 0.604 and 0.724 for ovarian and breast tumors, respectively. The RL agent achieved a peak Dice coefficient of 0.595 for ovarian tumors, outperforming random and alternative RL strategies. Additionally, it significantly reduced segmentation time, achieving a nearly ten-fold improvement compared to manual methods using SAM. Significance. While increased SAM prompts and non-centered prompts generally improved segmentation accuracy, each pathology and modality has specific optimal thresholds and placement strategies. Our RL agent achieved superior performance compared to other agents while achieving a significant reduction in segmentation time.

Fraction-variant VMAT planning for patients with complex gynecological and head-and-neck cancer

Abstract Objective . Increasing the number of arcs in volumetric modulated arc therapy (VMAT) allows for better intensity modulation and may improve plan quality. However, this leads to longer delivery times, which may cause patient discomfort and increase intra-fractional motion. In this study, it was investigated whether the delivery of different VMAT plans in different fractions may improve the dosimetric quality and delivery efficiency for the treatment of patients with complex tumor geometries. Approach . A direct aperture optimization algorithm was developed which allows for the simultaneous optimization of different VMAT plans to be delivered in different fractions, based on their cumulative physical dose. Each VMAT plan is constrained to deliver a uniform dose within the target volume, such that the entire treatment does not alter the fractionation scheme and is robust against inter-fractional setup errors. This approach was evaluated in-silico for five patients with gynecological and five patients with head-and-neck cancer. Main results . For all patients, fraction-variant treatments achieved significantly better target coverage and reduced the dose to critical organs-at-risk compared to fraction-invariant treatments that deliver the same plan in every fraction, where the dosimetric benefit was shown to increase with the number of different plans to be delivered. In addition, 1-arc and 2-arc fraction-variant treatments could approximate the dosimetric quality of 3-arc fraction-invariant treatments, while reducing the delivery time. Significance. Fraction-variant VMAT treatments may achieve excellent dosimetric quality for patients with complex tumor geometries, while keeping the delivery time per fraction viable.

GAN-based standardization of MR images: a promising approach for the development of multicentre radiomic models

Abstract Objective. This study evaluated generative adversarial network (GAN)-based magnetic resonance imaging (MRI) standardization methods by comparing them with conventional preprocessing and a posteriori approaches from the literature in their ability to mitigate the influence of acquisition parameters on radiomic analyses. Approach. MR T2-weighted images (T2w) of 30 patients with locally advanced cervical cancer (LACC) were acquired prospectively (cohort 1). For each patient, three images were taken sequentially on the same scanner with different values of repetition time (TR) and voxel size (VS). A retrospective cohort of 160 LACC patients (cohort 2) was also collected, including 86 and 160 T2w MR images taken before radiation therapy and brachytherapy, respectively. A conditional GAN (cGAN) and a cycleGAN were trained on cohort 1 and cohort 2, respectively, to generate images robust to the impact of acquisition parameters. This generative network-based standardization approach was compared to histogram-matching standardization, z-score standardization, and the ComBat harmonization methods. In this aim, different image quality metrics were extracted from cohort 1 images and the impact of standardization methods was assessed using principal component analysis (PCA). Using intra-class correlation (ICC) and concordance correlation coefficient (CCC), robust features were characterized (CCC&ICC ⩾ 0.75). Different classical ML models were finally trained to investigate the impact of these harmonization methods on stage classification and relapse prediction, respectively. Main results. PCA on quality metrics showed that the changes in TR and VS were mitigated the most with cGAN. Regarding TR/VS modulation, cGAN achieved the best results for the first- and second-order features, with 18/18 and 58/75 robust features, respectively. In both clinical tasks, ROC-AUC improved after standardization. For tumor stage classification, the application of a cycleGAN strategy significantly improved the performance of trained models compared to classification using raw images (ROC-AUC of 0.68 ± 0.16 before standardization and 0.88 ± 0.09 after standardization for the best ML model, i.e. logistic regression). Significance. GAN-based standardization in MRI could be an additional building block for robust radiomic signatures on a multicenter scale.

Automating IMRT planning for cervical cancer using dimension-scaled prior-based Vanilla Bayesian optimization

Abstract Objective. Manual inverse planning for radiation therapy is labor-intensive and often prone to inconsistent plan quality due to multiple adjustable planning hyperparameters, varying planner experience, and differing time constraints. Automated treatment planning offers a solution to these challenges. The present study investigates the effectiveness of dimension-scaled prior (DSP) Vanilla Bayesian optimization (BO) with a log-expected improvement (logEI) acquisition function in automating intensity-modulated radiation therapy (IMRT) planning for cancer cervix (CaCx) in high-dimensional settings. Approach. A Python-based auto-optimization script utilizing DSP Vanilla BO with logEI was employed to iteratively optimize the planning hyperparameters, including dose objectives and their corresponding weights for CaCx case IMRT plans on the Varian Eclipse treatment planning system (TPS) v18.0. This approach was assessed in 30 retrospectively selected pelvic node-positive CaCx cases, and the dosimetric parameters, based on EMBRACE-II protocol, were compared with plans generated from manual, Sparse Axis-Aligned Subspace BO (SAASBO), and stopping criterion-based DSP Vanilla BO. Main results. DSP Vanilla BO plans demonstrated superior dose conformity ( C I 95 % PT V 45 & C I 80 % PT V 45 ) and organs at risk (OAR) sparing ( V 40 Gy Bowel , V 30 Gy Bowel , V 40 Gy Bladder , V 40 Gy Rectum , D 0.01 % femoral   heads , D m e a n femoral   heads , and D mean kidneys ) compared to manual planning with significant improvements ( p < 0.05), while maintaining adequate clinical target coverage for CTV N , PTV 55 , ITV 45 , and PTV 45 . Compared to SAASBO, DSP Vanilla BO achieved comparable dosimetric quality but with less computation time (∼94 vs 360 min). The addition of stopping criteria further reduced the optimization time to ∼44 min while maintaining a plan quality comparable to manual planning. Significance. The study demonstrated that DSP Vanilla BO automated plans achieved comparable target coverage, with an improvement in OAR sparing, compared to manual plans. This highlights its effectiveness as an efficient, data-independent method for automating IMRT planning, which can be easily integrated into a TPS and benefit clinics with limited resources.

A feasibility study of automating radiotherapy planning with large language model agents

Abstract Objective. Radiotherapy planning requires significant expertise to balance tumor control and organ-at-risk (OAR) sparing. Automated planning can improve both efficiency and quality. This study introduces GPT-Plan, a novel multi-agent system powered by the GPT-4 family of large language models (LLMs), for automating the iterative radiotherapy plan optimization. Approach. GPT-Plan uses LLM-driven agents, mimicking the collaborative clinical workflow of a dosimetrist and physicist, to iteratively generate and evaluate text-based radiotherapy plans based on predefined criteria. Supporting tools assist the agents by leveraging historical plans, mitigating LLM hallucinations, and balancing exploration and exploitation. Performance was evaluated on 12 lung (IMRT) and 5 cervical (VMAT) cancer cases, benchmarked against the ECHO auto-planning method and manual plans. The impact of historical plan retrieval on efficiency was also assessed. Results. For IMRT lung cancer cases, GPT-Plan generated high-quality plans, demonstrating superior target coverage and homogeneity compared to ECHO while maintaining comparable or better OAR sparing. For VMAT cervical cancer cases, plan quality was comparable to a senior physicist and consistently superior to a junior physicist, particularly for OAR sparing. Retrieving historical plans significantly reduced the number of required optimization iterations for lung cases (p < 0.01) and yielded iteration counts comparable to those of the senior physicist for cervical cases (p = 0.313). Occasional LLM hallucinations have been mitigated by self-reflection mechanisms. One limitation was the inaccuracy of vision-based LLMs in interpreting dose images. Significance. This pioneering study demonstrates the feasibility of automating radiotherapy planning using LLM-powered agents for complex treatment decision-making tasks. While challenges remain in addressing LLM limitations, ongoing advancements hold potential for further refining and expanding GPT-Plan’s capabilities.

Dual convolution-transformer UNet (DCT-UNet) for organs at risk and clinical target volume segmentation in MRI for cervical cancer brachytherapy

Abstract Objective. MRI is the standard imaging modality for high-dose-rate brachytherapy of cervical cancer. Precise contouring of organs at risk (OARs) and high-risk clinical target volume (HR-CTV) from MRI is a crucial step for radiotherapy planning and treatment. However, conventional manual contouring has limitations in terms of accuracy as well as procedural time. To overcome these, we propose a deep learning approach to automatically segment OARs (bladder, rectum, and sigmoid colon) and HR-CTV from female pelvic MRI. Approach. In the proposed pipeline, a coarse multi-organ segmentation model first segments all structures, from which a region of interest is computed for each structure. Then, each organ is segmented using an organ-specific fine segmentation model separately trained for each organ. To account for variable sizes of HR-CTV, a size-adaptive multi-model approach was employed. For coarse and fine segmentations, we designed a dual convolution-transformer UNet (DCT-UNet) which uses dual-path encoder consisting of convolution and transformer blocks. To evaluate our model, OAR segmentations were compared to the clinical contours drawn by the attending radiation oncologist. For HR-CTV, four sets of contours (clinical + three additional sets) were obtained to produce a consensus ground truth as well as for inter/intra-observer variability analysis. Main results. DCT-UNet achieved dice similarity coefficient (mean ± SD) of 0.932 ± 0.032 (bladder), 0.786 ± 0.090 (rectum), 0.663 ± 0.180 (sigmoid colon), and 0.741 ± 0.076 (HR-CTV), outperforming other state-of-the-art models. Notably, the size-adaptive multi-model significantly improved HR-CTV segmentation compared to a single-model. Furthermore, significant inter/intra-observer variability was observed, and our model showed comparable performance to all observers. Computation time for the entire pipeline per subject was 12.59 ± 0.79 s, which is significantly shorter than the typical manual contouring time of >15 min. Significance. These experimental results demonstrate that our model has great utility in cervical cancer brachytherapy by enabling fast and accurate automatic segmentation, and has potential in improving consistency in contouring. DCT-UNet source code is available at https://github.com/JHU-MICA/DCT-UNet.

Deep learning-based segmentation for high-dose-rate brachytherapy in cervical cancer using 3D Prompt-ResUNet

Abstract Objective. To develop and evaluate a 3D Prompt-ResUNet module that utilized the prompt-based model combined with 3D nnUNet for rapid and consistent autosegmentation of high-risk clinical target volume (HRCTV) and organ at risk (OAR) in high-dose-rate brachytherapy for cervical cancer patients. Approach. We used 73 computed tomography scans and 62 magnetic resonance imaging scans from 135 (103 for training, 16 for validation, and 16 for testing) cervical cancer patients across two hospitals for HRCTV and OAR segmentation. A novel comparison of the deep learning neural networks 3D Prompt-ResUNet, nnUNet, and segment anything model-Med3D was applied for the segmentation. Evaluation was conducted in two parts: geometric and clinical assessments. Quantitative metrics included the Dice similarity coefficient (DSC), 95th percentile Hausdorff distance (HD95%), Jaccard index (JI), and Matthews correlation coefficient (MCC). Clinical evaluation involved interobserver comparison, 4-grade expert scoring, and a double-blinded Turing test. Main results. The Prompt-ResUNet model performed most similarly to experienced radiation oncologists, outperforming less experienced ones. During testing, the DSC, HD95% (mm), JI, and MCC value (mean ± SD) for HRCTV were 0.92 ± 0.03, 2.91 ± 0.69, 0.85 ± 0.04, and 0.92 ± 0.02, respectively. For the bladder, these values were 0.93 ± 0.05, 3.07 ± 1.05, 0.87 ± 0.08, and 0.93 ± 0.05, respectively. For the rectum, they were 0.87 ± 0.03, 3.54 ± 1.46, 0.78 ± 0.05, and 0.87 ± 0.03, respectively. For the sigmoid, they were 0.76 ± 0.11, 7.54 ± 5.54, 0.63 ± 0.14, and 0.78 ± 0.09, respectively. The Prompt-ResUNet achieved a clinical viability score of at least 2 in all evaluation cases (100%) for both HRCTV and bladder and exceeded the 30% positive rate benchmark for all evaluated structures in the Turing test. Significance. The Prompt-ResUNet architecture demonstrated high consistency with ground truth in autosegmentation of HRCTV and OARs, reducing interobserver variability and shortening treatment times.

An interactive nuclei segmentation framework with Voronoi diagrams and weighted convex difference for cervical cancer pathology images

Abstract Objective. Nuclei segmentation is crucial for pathologists to accurately classify and grade cancer. However, this process faces significant challenges, such as the complex background structures in pathological images, the high-density distribution of nuclei, and cell adhesion. Approach. In this paper, we present an interactive nuclei segmentation framework that increases the precision of nuclei segmentation. Our framework incorporates expert monitoring to gather as much prior information as possible and accurately segment complex nucleus images through limited pathologist interaction, where only a small portion of the nucleus locations in each image are labeled. The initial contour is determined by the Voronoi diagram generated from the labeled points, which is then input into an optimized weighted convex difference model to regularize partition boundaries in an image. Specifically, we provide theoretical proof of the mathematical model, stating that the objective function monotonically decreases. Furthermore, we explore a postprocessing stage that incorporates histograms, which are simple and easy to handle and prevent arbitrariness and subjectivity in individual choices. Main results. To evaluate our approach, we conduct experiments on both a cervical cancer dataset and a nasopharyngeal cancer dataset. The experimental results demonstrate that our approach achieves competitive performance compared to other methods. Significance. The Voronoi diagram in the paper serves as prior information for the active contour, providing positional information for individual cells. Moreover, the active contour model achieves precise segmentation results while offering mathematical interpretability.

Afterloader integrated EMT enables improved dwell position model definition and quality assurance in Venezia gynaecological brachytherapy applicators

Abstract Objective. In brachytherapy for gynecological cancers using intracavitary applicators, implant reconstruction is commonly performed using applicator libraries. These libraries contain applicator geometry models as well as dwell position (DP) models defined in respect to the applicator geometry. In this study, we investigate whether an afterloader integrated electromagnetic tracking (EMT) system can be utilized for DP model definition and quality assurance in such applicators. Approach. DPs in four sets of two configurations of the Elekta Venezia Advanced Gynaecological Applicator (22 mm ovoids/40 mm intrauterine (IU) and 26 mm ovoids/70 mm IU) were measured using an afterloader integrated EMT system. Measurements were evaluated for reproducibility and compared against manufacturer-specified (MS) DPs and a computed tomography (CT)-corrected DP model. Main Results. Excellent EMT measurement reproducibility was observed, with values of ⩽0.2 mm for both configurations. The overall reproducibility, including applicator geometry reproducibility, was ⩽0.4 mm for both configurations. Significant discrepancies from the manufacturer’s DP model were observed, with a mean ± sd deviation of 1.13 ± 0.66 mm (22/40) and 1.37 ± 0.63 (26/70), particularly in the IU channel, where MS DPs were not experimentally defined. Discrepancies were reduced to 0.89 ± 0.41 mm (22/40) and 0.81 ± 0.33 mm (26/70) when the CT-corrected DP model was used as baseline, highlighting the need for experimentally defined DP models. The overall uncertainty of single measurements was below the clinically acceptable 2 mm limit. Significance. This study confirms that afterloader integrated EMT can accurately reconstruct source paths in gynecological brachytherapy applicators and supports its incorporation into clinical workflows for improved quality assurance and treatment precision. The importance of EMT for quality assurance was highlighted by measured deviations from manufacturer’s DP model in a clinical relevant part of the IU channel.

Prediction of ovarian cancer prognosis using statistical radiomic features of ultrasound images

Abstract Objective. Ovarian cancer is the deadliest gynecologic malignancy worldwide. Ultrasound is the most useful non-invasive test for preoperative diagnosis of ovarian cancer. In this study, by leveraging multiple ultrasound images from the same patient to generate personalized, informative statistical radiomic features, we aimed to develop improved ultrasound image-based prognostic models for ovarian cancer. Approach. A total of 2057 ultrasound images from 514 ovarian cancer patients, including 355 patients with epithelial ovarian cancer, from two hospitals in China were collected for this study. The models were constructed using our recently developed Frequency Appearance in Multiple Univariate pre-Screening feature selection algorithm and Cox proportional hazards model. Main results. The models showed high predictive performance for overall survival (OS) and recurrence-free survival (RFS) in both epithelial and nonepithelial ovarian cancer, with concordance indices ranging from 0.773 to 0.794. Radiomic scores predicted 2 year OS and RFS risk groups with significant survival differences (log-rank test, P < 1.0 × 10−4 for both validation cohorts). OS and RFS hazard ratios between low- and high-risk groups were 15.994 and 30.692 (internal cohort) and 19.339 and 19.760 (external cohort), respectively. The improved performance of these newly developed prognostic models was mainly attributed to the use of multiple preoperative ultrasound images from the same patient to generate statistical radiomic features, rather than simply using the largest tumor region of interest among them. The models also revealed that the roundness of tumor lesion shape was positively correlated with prognosis for ovarian cancer. Significance. The newly developed prognostic models based on statistical radiomic features from ultrasound images were highly predictive of the risk of cancer-related death and possible recurrence not only for patients with epithelial ovarian cancer but also for those with nonepithelial ovarian cancer. They thereby provide reliable, non-invasive markers for individualized prognosis evaluation and clinical decision-making for patients with ovarian cancer.

Deformation trajectory prediction using a neural network trained on finite element data—application to library of CTVs creation for cervical cancer

Abstract Purpose . We propose a neural network for fast prediction of realistic, time-parametrized deformations between pairs of input segmentations. The proposed method was used to generate a library of planning CTVs for cervical cancer radiotherapy. Methods. A 3D convolutional neural network (CNN) was introduced to predict a stationary velocity field given the distance maps of the cervix CTV in empty and full bladder anatomy. Diffeomorphic deformation trajectories between the two states were obtained by time integration. Intermediate deformation states were used to populate a library of cervix CTVs. The network was trained on cervix CTV deformations of 20 patients generated by finite element modeling (FEM). Validation was performed on FEM data of 9 healthy volunteers. Additionally, for these subjects, CTV deformations were observed in a series of repeat MR scans as the bladder filled from empty to full. Predicted and FEM libraries were compared, and benchmarked against the observed deformations. Finally, for an independent test set of 20 patients the predicted libraries were evaluated clinically, and compared to the current method. Results. The median Dice score over the validation subjects between the predicted and FEM libraries was >0.95 throughout the deformation, with a median 90 percentile surface distance of <3 mm. The ability to cover observed CTVs was similar for both the FEM-based and the proposed method, with residual offsets being about twice as large as the difference between the two methods. Clinical evaluation showed improved library properties over the method currently used in clinic. Conclusions. We proposed a CNN trained on FEM deformations, which predicts the deformation trajectory between two input states of the cervix CTV in one forward pass. We applied this to CTV library prediction for cervical cancer. The network is able to mimic FEM deformations well, while being much faster and simpler in use.

A linear optimization model for high dose rate brachytherapy using a novel distance metric

Abstract Purpose. We propose a linear network-based optimization model (LNBM) for high dose rate brachytherapy (HDR-BT) that uses a novel distance metric to measure the discrepancy between the dose delivered and the prescription. Unlike models in the literature, LNBM takes advantage of the adjacency structure of the patients’ voxels by formalizing them into a network. Methods. We apply LNBM to a set of 7 cervical cancer cases treated with HDR-BT. State-of-the-art commercial optimization software solves LNBM to global optimality. The results of LNBM are compared with those of inverse planning by simulated annealing (IPSA) based on tumor coverage, dosimetric indices for the critical organs at risk (OARs), isodose contour plots, and two metrics of homogeneity new to this work (hot-spots volumes and diameters). Results. LNBM produces plans with improved tumor coverage and with improved isodose contour plots and dosimetric indices for OARs that receive highest dose (bladder and rectum in this study) when compared with IPSA. Using new metrics of homogeneity, we also demonstrate that LNBM produces more homogeneous plans on these cases. An analysis of the solutions of LNBM shows that they use a significant part of the voxel network structure, providing evidence that the plans produced are different from those created using traditional penalty approaches and are more directly guided by the geometry of the patients’ anatomy. Conclusions. The proposed linear network-based optimization model efficiently generates more homogeneous high quality treatment plans for HDR-BT.

Incorporating cross-voxel exchange for the analysis of dynamic contrast-enhanced imaging data: pre-clinical results

Abstract Tumours exhibit abnormal interstitial structures and vasculature function often leading to impaired and heterogeneous drug delivery. The disproportionate spatial accumulation of a drug in the interstitium is determined by several microenvironmental properties (blood vessel distribution and permeability, gradients in the interstitial fluid pressure). Predictions of tumour perfusion are key determinants of drug delivery and responsiveness to therapy. Pharmacokinetic models allow for the quantification of tracer perfusion based on contrast enhancement measured with non-invasive imaging techniques. An advanced cross-voxel exchange model (CVXM) was recently developed to provide a comprehensive description of tracer extravasation as well as advection and diffusion based on cross-voxel tracer kinetics (Sinno et al 2021). Transport parameters were derived from DCE-MRI of twenty TS-415 human cervical carcinoma xenografts by using CVXM. Tracer velocity flows were measured at the tumour periphery (mean 1.78–5.82 μm.s−1) pushing the contrast outward towards normal tissue. These elevated velocity measures and extravasation rates explain the heterogeneous distribution of tracer across the tumour and its accumulation at the periphery. Significant values for diffusivity were deduced across the tumours (mean 152–499 μm2.s−1). CVXM resulted in generally smaller values for the extravasation parameter K e x t (mean 0.01–0.04 min−1) and extravascular extracellular volume fraction v e (mean 0.05–0.17) compared to the standard Tofts parameters, suggesting that Toft model underestimates the effects of inter-voxel exchange. The ratio of Tofts’ extravasation parameters over CVXM’s was significantly positively correlated to the cross-voxel diffusivity (P < 0.0001) and velocity (P = 0.0005). Tofts’ increased v e measurements were explained using Sinno et al (2021)’s theoretical work. Finally, a scan time of 15 min renders informative estimations of the transport parameters. However, a duration as low as 7.5 min is acceptable to recognize the spatial variation of transport parameters. The results demonstrate the potential of utilizing CVXM for determining metrics characterizing the exchange of tracer between the vasculature and the tumour tissue. Like for many earlier models, additional work is strongly recommended, in terms of validation, to develop more confidence in the results, motivating future laboratory work in this regard.