Automatic Segmentation Using Deep Learning to Enable Online Dose Optimization During Adaptive Radiation Therapy of Cervical Cancer

Accurate object localization facilitates automatic esophagus segmentation in deep learning

BACKGROUND

Currently, automatic esophagus segmentation remains a challenging task due to its small size, low contrast, and large shape variation. We aimed to improve the performance of esophagus segmentation in deep learning by applying a strategy that involves locating the object first and then performing the segmentation task.

METHODS

A total of 100 cases with thoracic computed tomography scans from two publicly available datasets were used in this study. A modified CenterNet, an object location network, was employed to locate the center of the esophagus for each slice. Subsequently, the 3D U-net and 2D U-net_coarse models were trained to segment the esophagus based on the predicted object center. A 2D U-net_fine model was trained based on the updated object center according to the 3D U-net model. The dice similarity coefficient and the 95% Hausdorff distance were used as quantitative evaluation indexes for the delineation performance. The characteristics of the automatically delineated esophageal contours by the 2D U-net and 3D U-net models were summarized. Additionally, the impact of the accuracy of object localization on the delineation performance was analyzed. Finally, the delineation performance in different segments of the esophagus was also summarized.

RESULTS

The mean dice coefficient of the 3D U-net, 2D U-net_coarse, and 2D U-net_fine models were 0.77, 0.81, and 0.82, respectively. The 95% Hausdorff distance for the above models was 6.55, 3.57, and 3.76, respectively. Compared with the 2D U-net, the 3D U-net has a lower incidence of delineating wrong objects and a higher incidence of missing objects. After using the fine object center, the average dice coefficient was improved by 5.5% in the cases with a dice coefficient less than 0.75, while that value was only 0.3% in the cases with a dice coefficient greater than 0.75. The dice coefficients were lower for the esophagus between the orifice of the inferior and the pulmonary bifurcation compared with the other regions.

CONCLUSION

The 3D U-net model tended to delineate fewer incorrect objects but also miss more objects. Two-stage strategy with accurate object location could enhance the robustness of the segmentation model and significantly improve the esophageal delineation performance, especially for cases with poor delineation results.

NRG Oncology Assessment of Artificial Intelligence Deep Learning-Based Auto-segmentation for Radiation Therapy: Current Developments, Clinical Considerations, and Future Directions

Deep learning neural networks (DLNN) in Artificial intelligence (AI) have been extensively explored for automatic segmentation in radiotherapy (RT). In contrast to traditional model-based methods, data-driven AI-based models for auto-segmentation have shown high accuracy in early studies in research settings and controlled environment (single institution). Vendor-provided commercial AI models are made available as part of the integrated treatment planning system (TPS) or as a stand-alone tool that provides streamlined workflow interacting with the main TPS. These commercial tools have drawn clinics' attention thanks to their significant benefit in reducing the workload from manual contouring and shortening the duration of treatment planning. However, challenges occur when applying these commercial AI-based segmentation models to diverse clinical scenarios, particularly in uncontrolled environments. Contouring nomenclature and guideline standardization has been the main task undertaken by the NRG Oncology. AI auto-segmentation holds the potential clinical trial participants to reduce interobserver variations, nomenclature non-compliance, and contouring guideline deviations. Meanwhile, trial reviewers could use AI tools to verify contour accuracy and compliance of those submitted datasets. In recognizing the growing clinical utilization and potential of these commercial AI auto-segmentation tools, NRG Oncology has formed a working group to evaluate the clinical utilization and potential of commercial AI auto-segmentation tools. The group will assess in-house and commercially available AI models, evaluation metrics, clinical challenges, and limitations, as well as future developments in addressing these challenges. General recommendations are made in terms of the implementation of these commercial AI models, as well as precautions in recognizing the challenges and limitations.

The use of deep learning in interventional radiotherapy (brachytherapy): A review with a focus on open source and open data

Deep learning advanced to one of the most important technologies in almost all medical fields. Especially in areas, related to medical imaging it plays a big role. However, in interventional radiotherapy (brachytherapy) deep learning is still in an early phase. In this review, first, we investigated and scrutinised the role of deep learning in all processes of interventional radiotherapy and directly related fields. Additionally, we summarised the most recent developments. For better understanding, we provide explanations of key terms and approaches to solving common deep learning problems. To reproduce results of deep learning algorithms both source code and training data must be available. Therefore, a second focus of this work is on the analysis of the availability of open source, open data and open models. In our analysis, we were able to show that deep learning plays already a major role in some areas of interventional radiotherapy, but is still hardly present in others. Nevertheless, its impact is increasing with the years, partly self-propelled but also influenced by closely related fields. Open source, data and models are growing in number but are still scarce and unevenly distributed among different research groups. The reluctance in publishing code, data and models limits reproducibility and restricts evaluation to mono-institutional datasets. The conclusion of our analysis is that deep learning can positively change the workflow of interventional radiotherapy but there is still room for improvements when it comes to reproducible results and standardised evaluation methods.

Automatic segmentation of tumour and organs at risk in 3D MRI for cervical cancer radiation therapy with anatomical variations

Cervical cancer is a common cancer in women globally, with treatment usually involving radiation therapy (RT). Accurate segmentation for the tumour site and organ-at-risks (OARs) could assist in the reduction of treatment side effects and improve treatment planning efficiency. Cervical cancer Magnetic Resonance Imaging (MRI) segmentation is challenging due to a limited amount of training data available and large inter- and intra- patient shape variation for OARs. The proposed Masked-Net consists of a masked encoder within the 3D U-Net to account for the large shape variation within the dataset, with additional dilated layers added to improve segmentation performance. A new loss function was introduced to consider the bounding box loss during training with the proposed Masked-Net. Transfer learning from a male pelvis MRI data with a similar field of view was included. The approaches were compared to the 3D U-Net which was widely used in MRI image segmentation. The data used consisted of 52 volumes obtained from 23 patients with stage IB to IVB cervical cancer across a maximum of 7 weeks of RT with manually contoured labels including the bladder, cervix, gross tumour volume, uterus and rectum. The model was trained and tested with a 5-fold cross validation. Outcomes were evaluated based on the Dice Similarity Coefficients (DSC), the Hausdorff Distance (HD) and the Mean Surface Distance (MSD). The proposed method accounted for the small dataset, large variations in OAR shape and tumour sizes with an average DSC, HD and MSD for all anatomical structures of 0.790, 30.19mm and 3.15mm respectively.

Prospective Evaluation of Automated Contouring for CT-Based Brachytherapy for Gynecologic Malignancies

Purpose

The use of deep learning to auto-contour organs at risk (OARs) in gynecologic radiation treatment is well established. Yet, there is limited data investigating the prospective use of auto-contouring in clinical practice. In this study, we assess the accuracy and efficiency of auto-contouring OARs for computed tomography-based brachytherapy treatment planning of gynecologic malignancies.

Methods and Materials

An inhouse contouring tool automatically delineated 5 OARs in gynecologic radiation treatment planning: the bladder, small bowel, sigmoid, rectum, and urethra. Accuracy of each auto-contour was evaluated using a 5-point Likert scale: a score of 5 indicated the contour could be used without edits, while a score of 1 indicated the contour was unusable. During scoring, automated contours were edited and subsequently used for treatment planning. Dice similarity coefficient, mean surface distance, 95% Hausdorff distance, Hausdorff distance, and dosimetric changes between original and edited contours were calculated. Contour approval time and total planning time of a prospective auto-contoured (AC) cohort were compared with times from a retrospective manually contoured (MC) cohort.

Results

Thirty AC cases from January 2022 to July 2022 and 31 MC cases from July 2021 to January 2022 were included. The mean (±SD) Likert score for each OAR was the following: bladder 4.77 (±0.58), small bowel 3.96 (±0.91), sigmoid colon 3.92 (±0.81), rectum 4.6 (±0.71), and urethra 4.27 (±0.78). No ACs required major edits. All OARs had a mean Dice similarity coefficient > 0.86, mean surface distance < 0.48 mm, 95% Hausdorff distance < 3.2 mm, and Hausdorff distance < 10.32 mm between original and edited contours. There was no significant difference in dose-volume histogram metrics (D2.0 cc/D0.1 cc) between original and edited contours (P values > .05). The average time to plan approval in the AC cohort was 19% less than the MC cohort. (AC vs MC, 117.0 + 18.0 minutes vs 144.9 ± 64.5 minutes, P = .045).

Conclusions

Automated contouring is useful and accurate in clinical practice. Auto-contouring OARs streamlines radiation treatment workflows and decreases time required to design and approve gynecologic brachytherapy plans.

Fully automated deep learning based auto-contouring of liver segments and spleen on contrast-enhanced CT images

Manual delineation of liver segments on computed tomography (CT) images for primary/secondary liver cancer (LC) patients is time-intensive and prone to inter/intra-observer variability. Therefore, we developed a deep-learning-based model to auto-contour liver segments and spleen on contrast-enhanced CT (CECT) images. We trained two models using 3d patch-based attention U-Net ([Formula: see text] and 3d full resolution of nnU-Net ([Formula: see text] to determine the best architecture ([Formula: see text]. BA was used with vessels ([Formula: see text] and spleen ([Formula: see text] to assess the impact on segment contouring. Models were trained, validated, and tested on 160 ([Formula: see text]), 40 ([Formula: see text]), 33 ([Formula: see text]), 25 (CCH) and 20 (CPVE) CECT of LC patients. [Formula: see text] outperformed [Formula: see text] across all segments with median differences in Dice similarity coefficients (DSC) ranging 0.03-0.05 (p < 0.05). [Formula: see text], and [Formula: see text] were not statistically different (p > 0.05), however, both were slightly better than [Formula: see text] by DSC up to 0.02. The final model, [Formula: see text], showed a mean DSC of 0.89, 0.82, 0.88, 0.87, 0.96, and 0.95 for segments 1, 2, 3, 4, 5-8, and spleen, respectively on entire test sets. Qualitatively, more than 85% of cases showed a Likert score [Formula: see text] 3 on test sets. Our final model provides clinically acceptable contours of liver segments and spleen which are usable in treatment planning.

Beam mask and sliding window-facilitated deep learning-based accurate and efficient dose prediction for pencil beam scanning proton therapy

BACKGROUND

Accurate and efficient dose calculation is essential for on-line adaptive planning in proton therapy. Deep learning (DL) has shown promising dose prediction results in photon therapy. However, there is a scarcity of DL-based dose prediction methods specifically designed for proton therapy. Successful dose prediction method for proton therapy should account for more challenging dose prediction problems in pencil beam scanning proton therapy (PBSPT) due to its sensitivity to heterogeneities.

PURPOSE

To develop a DL-based PBSPT dose prediction workflow with high accuracy and balanced complexity to support on-line adaptive proton therapy clinical decision and subsequent replanning.

METHODS

PBSPT plans of 103 prostate cancer patients (93 for training and the other 10 for independent testing) and 83 lung cancer patients (73 for training and the other 10 for independent testing) previously treated at our institution were included in the study, each with computed tomography scans (CTs), structure sets, and plan doses calculated by the in-house developed Monte-Carlo dose engine (considered as the ground truth in the model training and testing). For the ablation study, we designed three experiments corresponding to the following three methods: (1) Experiment 1, the conventional region of interest (ROI) (composed of targets and organs-at-risk [OARs]) method. (2) Experiment 2, the beam mask (generated by raytracing of proton beams) method to improve proton dose prediction. (3) Experiment 3, the sliding window method for the model to focus on local details to further improve proton dose prediction. A fully connected 3D-Unet was adopted as the backbone. Dose volume histogram (DVH) indices, 3D Gamma passing rates with a criterion of 3%/3 mm/10%, and dice coefficients for the structures enclosed by the iso-dose lines between the predicted and the ground truth doses were used as the evaluation metrics. The calculation time for each proton dose prediction was recorded to evaluate the method's efficiency.

RESULTS

Compared to the conventional ROI method, the beam mask method improved the agreement of DVH indices for both targets and OARs and the sliding window method further improved the agreement of the DVH indices (for lung cancer, CTV D98 absolute deviation: 0.74 ± 0.18 vs. 0.57 ± 0.21 vs. 0.54 ± 0.15 Gy[RBE], ROI vs. beam mask vs. sliding window methods, respectively). For the 3D Gamma passing rates in the target, OARs, and BODY (outside target and OARs), the beam mask method improved the passing rates in these regions and the sliding window method further improved them (for prostate cancer, targets: 96.93% ± 0.53% vs. 98.88% ± 0.49% vs. 99.97% ± 0.07%, BODY: 86.88% ± 0.74% vs. 93.21% ± 0.56% vs. 95.17% ± 0.59%). A similar trend was also observed for the dice coefficients. This trend was especially remarkable for relatively low prescription isodose lines (for lung cancer, 10% isodose line dice: 0.871 ± 0.027 vs. 0.911 ± 0.023 vs. 0.927 ± 0.017). The dose predictions for all the testing cases were completed within 0.25 s.

CONCLUSIONS

An accurate and efficient deep learning-augmented proton dose prediction framework has been developed for PBSPT, which can predict accurate dose distributions not only inside but also outside ROI efficiently. The framework can potentially further reduce the initial planning and adaptive replanning workload in PBSPT.

Investigation of autosegmentation techniques on T2-weighted MRI for off-line dose reconstruction in MR-linac workflow for head and neck cancers

BACKGROUND

In order to accurately accumulate delivered dose for head and neck cancer patients treated with the Adapt to Position workflow on the 1.5T magnetic resonance imaging (MRI)-linear accelerator (MR-linac), the low-resolution T2-weighted MRIs used for daily setup must be segmented to enable reconstruction of the delivered dose at each fraction.

PURPOSE

In this pilot study, we evaluate various autosegmentation methods for head and neck organs at risk (OARs) on on-board setup MRIs from the MR-linac for off-line reconstruction of delivered dose.

METHODS

Seven OARs (parotid glands, submandibular glands, mandible, spinal cord, and brainstem) were contoured on 43 images by seven observers each. Ground truth contours were generated using a simultaneous truth and performance level estimation (STAPLE) algorithm. Twenty total autosegmentation methods were evaluated in ADMIRE: 1-9) atlas-based autosegmentation using a population atlas library (PAL) of 5/10/15 patients with STAPLE, patch fusion (PF), random forest (RF) for label fusion; 10-19) autosegmentation using images from a patient's 1-4 prior fractions (individualized patient prior [IPP]) using STAPLE/PF/RF; 20) deep learning (DL) (3D ResUNet trained on 43 ground truth structure sets plus 45 contoured by one observer). Execution time was measured for each method. Autosegmented structures were compared to ground truth structures using the Dice similarity coefficient, mean surface distance (MSD), Hausdorff distance (HD), and Jaccard index (JI). For each metric and OAR, performance was compared to the inter-observer variability using Dunn's test with control. Methods were compared pairwise using the Steel-Dwass test for each metric pooled across all OARs. Further dosimetric analysis was performed on three high-performing autosegmentation methods (DL, IPP with RF and 4 fractions [IPP_RF_4], IPP with 1 fraction [IPP_1]), and one low-performing (PAL with STAPLE and 5 atlases [PAL_ST_5]). For five patients, delivered doses from clinical plans were recalculated on setup images with ground truth and autosegmented structure sets. Differences in maximum and mean dose to each structure between the ground truth and autosegmented structures were calculated and correlated with geometric metrics.

RESULTS

DL and IPP methods performed best overall, all significantly outperforming inter-observer variability and with no significant difference between methods in pairwise comparison. PAL methods performed worst overall; most were not significantly different from the inter-observer variability or from each other. DL was the fastest method (33 s per case) and PAL methods the slowest (3.7-13.8 min per case). Execution time increased with a number of prior fractions/atlases for IPP and PAL. For DL, IPP_1, and IPP_RF_4, the majority (95%) of dose differences were within ± 250 cGy from ground truth, but outlier differences up to 785 cGy occurred. Dose differences were much higher for PAL_ST_5, with outlier differences up to 1920 cGy. Dose differences showed weak but significant correlations with all geometric metrics (R2 between 0.030 and 0.314).

CONCLUSIONS

The autosegmentation methods offering the best combination of performance and execution time are DL and IPP_1. Dose reconstruction on on-board T2-weighted MRIs is feasible with autosegmented structures with minimal dosimetric variation from ground truth, but contours should be visually inspected prior to dose reconstruction in an end-to-end dose accumulation workflow.

Validation of a deep-learning segmentation model for adult and pediatric head and neck radiotherapy in different patient positions

Background and purpose

Autocontouring for radiotherapy has the potential to significantly save time and reduce interobserver variability. We aimed to assess the performance of a commercial autocontouring model for head and neck (H&N) patients in eight orientations relevant to particle therapy with fixed beam lines, focusing on validation and implementation for routine clinical use.

Materials and methods

Autocontouring was performed on sixteen organs at risk (OARs) for 98 adult and pediatric patients with 137 H&N CT scans in eight orientations. A geometric comparison of the autocontours and manual segmentations was performed using the Hausdorff Distance 95th percentile, Dice Similarity Coefficient (DSC) and surface DSC and compared to interobserver variability where available. Additional qualitative scoring and dose-volume-histogram (DVH) parameters analyses were performed for twenty patients in two positions, consisting of scoring on a 0-3 scale based on clinical usability and comparing the mean (Dmean) and near-maximum (D2%) dose, respectively.

Results

For the geometric analysis, the model performance in head-first-supine straight and hyperextended orientations was in the same range as the interobserver variability. HD95, DSC and surface DSC was heterogeneous in other orientations. No significant geometric differences were found between pediatric and adult autocontours. The qualitative scoring yielded a median score of ≥ 2 for 13/16 OARs while 7/32 DVH parameters were significantly different.

Conclusions

For head-first-supine straight and hyperextended scans, we found that 13/16 OAR autocontours were suited for use in daily clinical practice and subsequently implemented. Further development is needed for other patient orientations before implementation.

Script-based automatic radiotherapy planning for cervical cancer

BACKGROUND

This study aimed to develop fully automated script-based radiotherapy treatment plans for cervical cancer patients, and evaluate them against clinically accepted plans, as validation before clinical implementation.

MATERIAL AND METHODS

In this retrospective planning study, treatment plans for 25 locally advanced cervical cancer (LACC) patients with up to three dose levels were included. Fully automated plans were created using an in-house developed Python script in RayStation, and compared to clinically accepted manually made plans. Quantitatively, relevant dose statistics were compared, and average dose volume histograms (DVHs) were analyzed. Qualitatively, a blinded plan comparison was conducted between the clinical and automatic plans. The accuracy of treatment plan delivery was verified with the Delta4 Phantom+.

RESULTS

The quantitative evaluation showed that target coverage was acceptable for all the automatic and clinical plans. The automatic plans were significantly more conformal than the clinical plans; median of 1.03 vs. 1.12. Mean doses to almost all organs at risk (OARs) were reduced in the automatic plans, with a median reduction of between 0.6 Gy and 1.9 Gy. In the blinded plan comparison, the automatic plans were the preferred plans or of equal quality as the clinical plans in 99% of the cases. In addition, plan delivery was excellent, with a mean gamma passing rate of 99.8%. Complete script-based plans were generated in 30-45 min; about four to ten times faster than manually made plans.

CONCLUSION

The automatic plans had acceptable target coverage, lower doses to almost all OARs, more conformal dose distributions, and were predominantly preferred by the clinicians. Based on these results, our institution has implemented the script for clinical use.

Single patient learning for adaptive radiotherapy dose prediction

BACKGROUND

Throughout a patient's course of radiation therapy, maintaining accuracy of their initial treatment plan over time is challenging due to anatomical changes-for example, stemming from patient weight loss or tumor shrinkage. Online adaptation of their RT plan to these changes is crucial, but hindered by manual and time-consuming processes. While deep learning (DL) based solutions have shown promise in streamlining adaptive radiation therapy (ART) workflows, they often require large and extensive datasets to train population-based models.

PURPOSE

This study extends our prior research by introducing a minimalist approach to patient-specific adaptive dose prediction. In contrast to our prior method, which involved fine-tuning a pre-trained population model, this new method trains a model from scratch using only a patient's initial treatment data. This patient-specific dose predictor aims to enhance clinical accessibility, thereby empowering physicians and treatment planners to make more informed, quantitative decisions in ART. We hypothesize that patient-specific DL models will provide more accurate adaptive dose predictions for their respective patients compared to a population-based DL model.

METHODS

We selected 33 patients to train an adaptive population-based (AP) model. Ten additional patients were selected, and their respective initial RT data served as single samples for training patient-specific (PS) models. These 10 patients contained an additional 26 ART plans that were withheld as the test dataset to evaluate AP versus PS model dose prediction performance. We assessed model performance using Mean Absolute Percent Error (MAPE) by comparing predicted doses to the originally delivered ground truth doses. We used the Wilcoxon signed-rank test to determine statistically significant differences in terms of MAPE between the AP and PS model results across the test dataset. Furthermore, we calculated differences between predicted and ground truth mean doses for segmented structures and determined statistical significance in the differences for each of them.

RESULTS

The average MAPE across AP and PS model dose predictions was 5.759% and 4.069%, respectively. The Wilcoxon signed-rank test yielded two-tailed p-value = 2.9802 × 10 - 8 $2.9802\ \times \ {10}^{ - 8}$ , indicating that the MAPE differences between the AP and PS model dose predictions are statistically significant, and 95% confidence interval = [-2.1610, -1.0130], indicating 95% confidence that the MAPE difference between the AP and PS models for a population lies in this range. Out of 24 total segmented structures, the comparison of mean dose differences for 12 structures indicated statistical significance with two-tailed p-values < 0.05.

CONCLUSION

Our study demonstrates the potential of patient-specific deep learning models in application to ART. Notably, our method streamlines the training process by minimizing the size of the required training dataset, as only a single patient's initial treatment data is required. External institutions considering the implementation of such a technology could package such a model so that it only requires the upload of a reference treatment plan for model training and deployment. Our single patient learning strategy demonstrates promise in ART due to its minimal dataset requirement and its utility in personalization of cancer treatment.

Deep Learning Framework with Multi-Head Dilated Encoders for Enhanced Segmentation of Cervical Cancer on Multiparametric Magnetic Resonance Imaging

T2-weighted magnetic resonance imaging (MRI) and diffusion-weighted imaging (DWI) are essential components of cervical cancer diagnosis. However, combining these channels for the training of deep learning models is challenging due to image misalignment. Here, we propose a novel multi-head framework that uses dilated convolutions and shared residual connections for the separate encoding of multiparametric MRI images. We employ a residual U-Net model as a baseline, and perform a series of architectural experiments to evaluate the tumor segmentation performance based on multiparametric input channels and different feature encoding configurations. All experiments were performed on a cohort of 207 patients with locally advanced cervical cancer. Our proposed multi-head model using separate dilated encoding for T2W MRI and combined b1000 DWI and apparent diffusion coefficient (ADC) maps achieved the best median Dice similarity coefficient (DSC) score, 0.823 (confidence interval (CI), 0.595-0.797), outperforming the conventional multi-channel model, DSC 0.788 (95% CI, 0.568-0.776), although the difference was not statistically significant (p > 0.05). We investigated channel sensitivity using 3D GRAD-CAM and channel dropout, and highlighted the critical importance of T2W and ADC channels for accurate tumor segmentation. However, our results showed that b1000 DWI had a minor impact on the overall segmentation performance. We demonstrated that the use of separate dilated feature extractors and independent contextual learning improved the model's ability to reduce the boundary effects and distortion of DWI, leading to improved segmentation performance. Our findings could have significant implications for the development of robust and generalizable models that can extend to other multi-modal segmentation applications.

Artificial intelligence-based risk stratification, accurate diagnosis and treatment prediction in gynecologic oncology

As data-driven science, artificial intelligence (AI) has paved a promising path toward an evolving health system teeming with thrilling opportunities for precision oncology. Notwithstanding the tremendous success of oncological AI in such fields as lung carcinoma, breast tumor and brain malignancy, less attention has been devoted to investigating the influence of AI on gynecologic oncology. Hereby, this review sheds light on the ever-increasing contribution of state-of-the-art AI techniques to the refined risk stratification and whole-course management of patients with gynecologic tumors, in particular, cervical, ovarian and endometrial cancer, centering on information and features extracted from clinical data (electronic health records), cancer imaging including radiological imaging, colposcopic images, cytological and histopathological digital images, and molecular profiling (genomics, transcriptomics, metabolomics and so forth). However, there are still noteworthy challenges beyond performance validation. Thus, this work further describes the limitations and challenges faced in the real-word implementation of AI models, as well as potential solutions to address these issues.

Quantitative assessment of adaptive radiotherapy for prostate cancer using deep learning: Bladder dose as a decision criterion

BACKGROUND

Adaptive radiotherapy (ART) can incorporate anatomical variations in a reoptimized treatment plan for fractionated radiotherapy. An automatic solution to objectively determine whether ART should be performed immediately after the daily image acquisition is highly desirable.

PURPOSE

We investigate a quantitative criterion for whether ART should be performed in prostate cancer radiotherapy by synthesizing pseudo-CT (sCT) images and evaluating dosimetric impact on treatment planning using deep learning approaches.

METHOD AND MATERIALS

Planning CT (pCT) and daily cone-beam CT (CBCT) data sets of 74 patients are used to train (60 patients) and evaluate (14 patients) a cycle adversarial generative network (CycleGAN) that performs the task of synthesizing high-quality sCT from daily CBCT. Automatic delineation (AD) of the bladder is performed on the sCT using the U-net. The combination of sCT and AD allows us to perform dose calculations based on the up-to-date bladder anatomy to determine whether the original treatment plan (ori-plan) is still applicable. For positive cases that the patients' anatomical changes and the associated dose calculations warrant re-planning, we made rapid plan revisions (re-plan) based on the ori-plan.

RESULTS

The mean absolute error within the region-of-interests (i.e., body, bladder, fat, muscle) between the sCT and pCT are 41.2, 25.1, 26.5, and 29.0HU, respectively. Taking the calculated results of pCT doses as the standard, for PTV, the gamma passing rates of sCT doses at 1 mm/1%, 2 mm/2% are 87.92%, 98.78%, respectively. The Dice coefficients of the AD-contours are 0.93 on pCT and 0.91 on sCT. According to the result of dose calculation, we found when the bladder volume underwent a substantial change (79.7%), the bladder dose is still within the safe limit, suggesting it is insufficient to solely use the bladder volume change as a criterion to determine whether adaptive treatment needs to be done. After AD-contours of the bladder using sCT, there are two cases whose bladder doseD mean > 4000 cGy ${{\mathrm{D}}}_{{\mathrm{mean}}} > 4000{\mathrm{\ cGy}}$ . For the two cases, we perform re-planning to reduce the bladder dose toD mean = 3841 cGy ${{\mathrm{D}}}_{{\mathrm{mean}}} = 3841{\mathrm{\ cGy}}$ ,D mean = 3580 cGy ${{\mathrm{D}}}_{{\mathrm{mean}}} = 3580{\mathrm{\ cGy\ }}$ under the condition that the PTV meets the prescribed dose.

CONCLUSION

We provide a dose accurate adaptive workflow for prostate cancer patients by using deep learning approaches, and implement ART that adapts to bladder dose. Of note, the specific replanning criterion for whether ART needs to be performed can adapt to different centers' choices based on their experience and daily observations.

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

PURPOSE

Delineation of the clinical target volume (CTV) and organs-at-risk (OARs) is important in cervical cancer radiotherapy. But it is generally labor-intensive, time-consuming, and subjective. This paper proposes a parallel-path attention fusion network (PPAF-net) to overcome these disadvantages in the delineation task.

METHODS

The PPAF-net utilizes both the texture and structure information of CTV and OARs by employing a U-Net network to capture the high-level texture information, and an up-sampling and down-sampling (USDS) network to capture the low-level structure information to accentuate the boundaries of CTV and OARs. Multi-level features extracted from both networks are then fused together through an attention module to generate the delineation result.

RESULTS

The dataset contains 276 computed tomography (CT) scans of patients with cervical cancer of staging IB-IIA. The images are provided by the West China Hospital of Sichuan University. Simulation results demonstrate that PPAF-net performs favorably on the delineation of the CTV and OARs (e.g., rectum, bladder and etc.) and achieves the state-of-the-art delineation accuracy, respectively, for the CTV and OARs. In terms of the Dice Similarity Coefficient (DSC) and the Hausdorff Distance (HD), 88.61% and 2.25 cm for the CTV, 92.27% and 0.73 cm for the rectum, 96.74% and 0.68 cm for the bladder, 96.38% and 0.65 cm for the left kidney, 96.79% and 0.63 cm for the right kidney, 93.42% and 0.52 cm for the left femoral head, 93.69% and 0.51 cm for the right femoral head, 87.53% and 1.07 cm for the small intestine, and 91.50% and 0.84 cm for the spinal cord.

CONCLUSIONS

The proposed automatic delineation network PPAF-net performs well on CTV and OARs segmentation tasks, which has great potential for reducing the burden of radiation oncologists and increasing the accuracy of delineation. In future, radiation oncologists from the West China Hospital of Sichuan University will further evaluate the results of network delineation, making this method helpful in clinical practice.

Harnessing progress in radiotherapy for global cancer control

The pace of technological innovation over the past three decades has transformed the field of radiotherapy into one of the most technologically intense disciplines in medicine. However, the global barriers to access this highly effective treatment are complex and extend beyond technological limitations. Here, we review the technological advancement and current status of radiotherapy and discuss the efforts of the global radiation oncology community to formulate a more integrative 'diagonal approach' in which the agendas of science-driven advances in individual outcomes and the sociotechnological task of global cancer control can be aligned to bring the benefit of this proven therapy to patients with cancer everywhere.

Radiomics-based T-staging of hollow organ cancers

Cancer growing in hollow organs has become a serious threat to human health. The accurate T-staging of hollow organ cancers is a major concern in the clinic. With the rapid development of medical imaging technologies, radiomics has become a reliable tool of T-staging. Due to similar growth characteristics of hollow organ cancers, radiomics studies of these cancers can be used as a common reference. In radiomics, feature-based and deep learning-based methods are two critical research focuses. Therefore, we review feature-based and deep learning-based T-staging methods in this paper. In conclusion, existing radiomics studies may underestimate the hollow organ wall during segmentation and the depth of invasion in staging. It is expected that this survey could provide promising directions for following research in this realm.

Contour-guided deep learning based deformable image registration for dose monitoring during CBCT-guided radiotherapy of prostate cancer

PURPOSE

To evaluate deep learning (DL)-based deformable image registration (DIR) for dose accumulation during radiotherapy of prostate cancer patients.

METHODS AND MATERIALS

Data including 341 CBCTs (209 daily, 132 weekly) and 23 planning CTs from 23 patients was retrospectively analyzed. Anatomical deformation during treatment was estimated using free-form deformation (FFD) method from Elastix and DL-based VoxelMorph approaches. The VoxelMorph method was investigated using anatomical scans (VMorph_Sc) or label images (VMorph_Msk), or the combination of both (VMorph_Sc_Msk). Accumulated doses were compared with the planning dose.

RESULTS

The DSC ranges, averaged for prostate, rectum and bladder, were 0.60-0.71, 0.67-0.79, 0.93-0.98, and 0.89-0.96 for the FFD, VMorph_Sc, VMorph_Msk, and VMorph_Sc_Msk methods, respectively. When including both anatomical and label images, VoxelMorph estimated more complex deformations resulting in heterogeneous determinant of Jacobian and higher percentage of deformation vector field (DVF) folding (up to a mean value of 1.90% in the prostate). Large differences were observed between DL-based methods regarding estimation of the accumulated dose, showing systematic overdosage and underdosage of the bladder and rectum, respectively. The difference between planned mean dose and accumulated mean dose with VMorph_Sc_Msk reached a median value of +6.3 Gy for the bladder and -5.1 Gy for the rectum.

CONCLUSION

The estimation of the deformations using DL-based approach is feasible for male pelvic anatomy but requires the inclusion of anatomical contours to improve organ correspondence. High variability in the estimation of the accumulated dose depending on the deformable strategy suggests further investigation of DL-based techniques before clinical deployment.

Artificial intelligence applications in brachytherapy: A literature review

PURPOSE

Artificial intelligence (AI) has the potential to simplify and optimize various steps of the brachytherapy workflow, and this literature review aims to provide an overview of the work done in this field.

METHODS AND MATERIALS

We conducted a literature search in June 2022 on PubMed, Embase, and Cochrane for papers that proposed AI applications in brachytherapy.

RESULTS

A total of 80 papers satisfied inclusion/exclusion criteria. These papers were categorized as follows: segmentation (24), registration and image processing (6), preplanning (13), dose prediction and treatment planning (11), applicator/catheter/needle reconstruction (16), and quality assurance (10). AI techniques ranged from classical models such as support vector machines and decision tree-based learning to newer techniques such as U-Net and deep reinforcement learning, and were applied to facilitate small steps of a process (e.g., optimizing applicator selection) or even automate the entire step of the workflow (e.g., end-to-end preplanning). Many of these algorithms demonstrated human-level performance and offer significant improvements in speed.

CONCLUSIONS

AI has potential to augment, automate, and/or accelerate many steps of the brachytherapy workflow. We recommend that future studies adhere to standard reporting guidelines. We also stress the importance of using larger sample sizes and reporting results using clinically interpretable measures.

Patient-specific three-dimensional dose distribution prediction via deep learning for prostate cancer therapy: Improvement with the structure loss

PURPOSE

Deep learning (DL)-based dose distribution prediction can potentially reduce the cost of inverse planning process. We developed and introduced a structure-focused loss (L_struct) for 3D dose prediction to improve prediction accuracy. This study investigated the influence of L_struct on DL-based dose prediction for patients with prostate cancer. The proposed L_struct, which is similar in concept to dose-volume histogram (DVH)-based optimization in clinical practice, has the potential to provide more interpretable and accurate DL-based optimization.

METHODS

This study involved 104 patients who underwent prostate radiotherapy. We used 3D U-Net-based architecture to predict dose distributions from computed tomography and contours of the planning target volume and organs-at-risk. We trained two models using different loss functions: L2 loss and L_struct. Predicted doses were compared in terms of dose-volume parameters and the Dice similarity coefficient of isodose volume.

RESULTS

DVH analysis showed that the L_struct model had smaller errors from the ground truth than the L2 model. The L_struct model achieved more consistent dose distributions than the L2 model, with errors close to zero. The isodose Dice score of the L_struct model was greater than that of the L2 model by >20% of the prescribed dose.

CONCLUSIONS

We developed L_struct using labels of inputted contours for DL-based dose prediction for prostate radiotherapy. L_struct can be generalized to any DL architecture, thereby enhancing the dose prediction accuracy.

Automated Contouring and Planning in Radiation Therapy: What Is 'Clinically Acceptable'?

Developers and users of artificial-intelligence-based tools for automatic contouring and treatment planning in radiotherapy are expected to assess clinical acceptability of these tools. However, what is 'clinical acceptability'? Quantitative and qualitative approaches have been used to assess this ill-defined concept, all of which have advantages and disadvantages or limitations. The approach chosen may depend on the goal of the study as well as on available resources. In this paper, we discuss various aspects of 'clinical acceptability' and how they can move us toward a standard for defining clinical acceptability of new autocontouring and planning tools.

A Review of the Metrics Used to Assess Auto-Contouring Systems in Radiotherapy

Auto-contouring could revolutionise future planning of radiotherapy treatment. The lack of consensus on how to assess and validate auto-contouring systems currently limits clinical use. This review formally quantifies the assessment metrics used in studies published during one calendar year and assesses the need for standardised practice. A PubMed literature search was undertaken for papers evaluating radiotherapy auto-contouring published during 2021. Papers were assessed for types of metric and the methodology used to generate ground-truth comparators. Our PubMed search identified 212 studies, of which 117 met the criteria for clinical review. Geometric assessment metrics were used in 116 of 117 studies (99.1%). This includes the Dice Similarity Coefficient used in 113 (96.6%) studies. Clinically relevant metrics, such as qualitative, dosimetric and time-saving metrics, were less frequently used in 22 (18.8%), 27 (23.1%) and 18 (15.4%) of 117 studies, respectively. There was heterogeneity within each category of metric. Over 90 different names for geometric measures were used. Methods for qualitative assessment were different in all but two papers. Variation existed in the methods used to generate radiotherapy plans for dosimetric assessment. Consideration of editing time was only given in 11 (9.4%) papers. A single manual contour as a ground-truth comparator was used in 65 (55.6%) studies. Only 31 (26.5%) studies compared auto-contours to usual inter- and/or intra-observer variation. In conclusion, significant variation exists in how research papers currently assess the accuracy of automatically generated contours. Geometric measures are the most popular, however their clinical utility is unknown. There is heterogeneity in the methods used to perform clinical assessment. Considering the different stages of system implementation may provide a framework to decide the most appropriate metrics. This analysis supports the need for a consensus on the clinical implementation of auto-contouring.

Effect of bladder filling status on positioning errors in post-hysterectomy cervical cancer radiotherapy

Objective: To investigate the effect of different bladder filling states on positioning errors in radiotherapy for cervical cancer and obtain the reference range of bladder filling consistency during radiotherapy.Methods: Patients who underwent postoperative radiotherapy for cervical cancer in Nantong Tumor Hospital from October 2018 to December 2019 were selected. According to the bladder filling deviation, they were divided into group A1 (deviation < 20%) and group B1 (deviation ≥ 20%). The bladder filling variations of the two groups were compared with different positioning errors. Group A2 has a positioning error of <0.4 cm, and group B2 has a positioning error of ≥0.4 cm. The reference range of bladder filling consistency during radiotherapy is obtained by analyzing the composition ratio of different positioning errors of bladder filling deviation.Results: This study included 195 patients with cervical cancer. The error of longitudinal and vertical position in group B1 was significantly higher than that in group A1 (0.50 ± 0.34 vs. 0.26 ± 0.22 cm, p < 0.001, and 0.22 ± 0.17 vs. 0.16 ± 0.12 cm, p < 0.001). Compared with group B2, the absolute deviation of bladder filling in group A2 (54.1% ± 54.4% vs. 25.6% ± 22.7%, p < 0.001) was slight. The chi-square test showed significant differences in the proportion of the positioning state of different bladder filling forms (χ2 = 31.006, p < 0.001). In addition, there was a significant difference in the proportion of stability errors in patients with poor stability in different directions (χ2 = 118.551, p < 0.001).Conclusion: In patients with cervical cancer fixed in the supine position, a bladder capacity deviation <20% is easier to achieve excellent positioning with, and it can better control the positioning error of radiotherapy and ensure the positioning accuracy of dose distribution to the target area. It can also achieve good tumor treatment effects. This range can be used as a reference for bladder filling consistency in patients with cervical cancer undergoing radiotherapy.

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

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.

Deep learning in CT image segmentation of cervical cancer: a systematic review and meta-analysis

Background

This paper attempts to conduct a systematic review and meta-analysis of deep learning (DLs) models for cervical cancer CT image segmentation.

Methods

Relevant studies were systematically searched in PubMed, Embase, The Cochrane Library, and Web of science. The literature on DLs for cervical cancer CT image segmentation were included, a meta-analysis was performed on the dice similarity coefficient (DSC) of the segmentation results of the included DLs models. We also did subgroup analyses according to the size of the sample, type of segmentation (i.e., two dimensions and three dimensions), and three organs at risk (i.e., bladder, rectum, and femur). This study was registered in PROSPERO prior to initiation (CRD42022307071).

Results

A total of 1893 articles were retrieved and 14 articles were included in the meta-analysis. The pooled effect of DSC score of clinical target volume (CTV), bladder, rectum, femoral head were 0.86(95%CI 0.84 to 0.87), 0.91(95%CI 0.89 to 0.93), 0.83(95%CI 0.79 to 0.88), and 0.92(95%CI 0.91to 0.94), respectively. For the performance of segmented CTV by two dimensions (2D) and three dimensions (3D) model, the DSC score value for 2D model was 0.87 (95%CI 0.85 to 0.90), while the DSC score for 3D model was 0.85 (95%CI 0.82 to 0.87). As for the effect of the capacity of sample on segmentation performance, no matter whether the sample size is divided into two groups: greater than 100 and less than 100, or greater than 150 and less than 150, the results show no difference (P > 0.05). Four papers reported the time for segmentation from 15 s to 2 min.

Conclusion

DLs have good accuracy in automatic segmentation of CT images of cervical cancer with a less time consuming and have good prospects for future radiotherapy applications, but still need public high-quality databases and large-scale research verification.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13014-022-02148-6.

Leveraging deep learning-based segmentation and contours-driven deformable registration for dose accumulation in abdominal structures

PURPOSE

Discrepancies between planned and delivered dose to GI structures during radiation therapy (RT) of liver cancer may hamper the prediction of treatment outcomes. The purpose of this study is to develop a streamlined workflow for dose accumulation in a treatment planning system (TPS) during liver image-guided RT and to assess its accuracy when using different deformable image registration (DIR) algorithms.

MATERIALS AND METHODS

Fifty-six patients with primary and metastatic liver cancer treated with external beam radiotherapy guided by daily CT-on-rails (CTOR) were retrospectively analyzed. The liver, stomach and duodenum contours were auto-segmented on all planning CTs and daily CTORs using deep-learning methods. Dose accumulation was performed for each patient using scripting functionalities of the TPS and considering three available DIR algorithms based on: (i) image intensities only; (ii) intensities + contours; (iii) a biomechanical model (contours only). Planned and accumulated doses were converted to equivalent dose in 2Gy (EQD2) and normal tissue complication probabilities (NTCP) were calculated for the stomach and duodenum. Dosimetric indexes for the normal liver, GTV, stomach and duodenum and the NTCP values were exported from the TPS for analysis of the discrepancies between planned and the different accumulated doses.

RESULTS

Deep learning segmentation of the stomach and duodenum enabled considerable acceleration of the dose accumulation process for the 56 patients. Differences between accumulated and planned doses were analyzed considering the 3 DIR methods. For the normal liver, stomach and duodenum, the distribution of the 56 differences in maximum doses (D2%) presented a significantly higher variance when a contour-driven DIR method was used instead of the intensity only-based method. Comparing the two contour-driven DIR methods, differences in accumulated minimum doses (D98%) in the GTV were >2Gy for 15 (27%) of the patients. Considering accumulated dose instead of planned dose in standard NTCP models of the duodenum demonstrated a high sensitivity of the duodenum toxicity risk to these dose discrepancies, whereas smaller variations were observed for the stomach.

CONCLUSION

This study demonstrated a successful implementation of an automatic workflow for dose accumulation during liver cancer RT in a commercial TPS. The use of contour-driven DIR methods led to larger discrepancies between planned and accumulated doses in comparison to using an intensity only based DIR method, suggesting a better capability of these approaches in estimating complex deformations of the GI organs.

Using Artificial Intelligence for Optimization of the Processes and Resource Utilization in Radiotherapy

The radiotherapy (RT) process from planning to treatment delivery is a multistep, complex operation involving numerous levels of human-machine interaction and requiring high precision. These steps are labor-intensive and time-consuming and require meticulous coordination between professionals with diverse expertise. We reviewed and summarized the current status and prospects of artificial intelligence and machine learning relevant to the various steps in RT treatment planning and delivery workflow specifically in low- and middle-income countries (LMICs). We also searched the PubMed database using the search terms (Artificial Intelligence OR Machine Learning OR Deep Learning OR Automation OR knowledge-based planning AND Radiotherapy) AND (list of Low- and Middle-Income Countries as defined by the World Bank at the time of writing this review). The search yielded a total of 90 results, of which results with first authors from the LMICs were chosen. The reference lists of retrieved articles were also reviewed to search for more studies. No language restrictions were imposed. A total of 20 research items with unique study objectives conducted with the aim of enhancing RT processes were examined in detail. Artificial intelligence and machine learning can improve the overall efficiency of RT processes by reducing human intervention, aiding decision making, and efficiently executing lengthy, repetitive tasks. This improvement could permit the radiation oncologist to redistribute resources and focus on responsibilities such as patient counseling, education, and research, especially in resource-constrained LMICs.

A Survey on Deep Learning for Precision Oncology

Precision oncology, which ensures optimized cancer treatment tailored to the unique biology of a patient’s disease, has rapidly developed and is of great clinical importance. Deep learning has become the main method for precision oncology. This paper summarizes the recent deep-learning approaches relevant to precision oncology and reviews over 150 articles within the last six years. First, we survey the deep-learning approaches categorized by various precision oncology tasks, including the estimation of dose distribution for treatment planning, survival analysis and risk estimation after treatment, prediction of treatment response, and patient selection for treatment planning. Secondly, we provide an overview of the studies per anatomical area, including the brain, bladder, breast, bone, cervix, esophagus, gastric, head and neck, kidneys, liver, lung, pancreas, pelvis, prostate, and rectum. Finally, we highlight the challenges and discuss potential solutions for future research directions.

A Feasibility Study of Deep Learning-Based Auto-Segmentation Directly Used in VMAT Planning Design and Optimization for Cervical Cancer

Purpose

To investigate the dosimetric impact on target volumes and organs at risk (OARs) when unmodified auto-segmented OAR contours are directly used in the design of treatment plans.

Materials and Methods

A total of 127 patients with cervical cancer were collected for retrospective analysis, including 105 patients in the training set and 22 patients in the testing set. The 3D U-net architecture was used for model training and auto-segmentation of nine types of organs at risk. The auto-segmented and manually segmented organ contours were used for treatment plan optimization to obtain the AS-VMAT (automatic segmentations VMAT) plan and the MS-VMAT (manual segmentations VMAT) plan, respectively. Geometric accuracy between the manual and predicted contours were evaluated using the Dice similarity coefficient (DSC), mean distance-to-agreement (MDA), and Hausdorff distance (HD). The dose volume histogram (DVH) and the gamma passing rate were used to identify the dose differences between the AS-VMAT plan and the MS-VMAT plan.

Results

Average DSC, MDA and HD₉₅ across all OARs were 0.82–0.96, 0.45–3.21 mm, and 2.30–17.31 mm on the testing set, respectively. The D_99% in the rectum and the Dmean in the spinal cord were 6.04 Gy (P = 0.037) and 0.54 Gy (P = 0.026) higher, respectively, in the AS-VMAT plans than in the MS-VMAT plans. The V₂₀, V₃₀, and V₄₀ in the rectum increased by 1.35% (P = 0.027), 1.73% (P = 0.021), and 1.96% (P = 0.008), respectively, whereas the V₁₀ in the spinal cord increased by 1.93% (P = 0.011). The differences in other dosimetry parameters were not statistically significant. The gamma passing rates in the clinical target volume (CTV) were 92.72% and 98.77%, respectively, using the 2%/2 mm and 3%/3 mm criteria, which satisfied the clinical requirements.

Conclusions

The dose distributions of target volumes were unaffected when auto-segmented organ contours were used in the design of treatment plans, whereas the impact of automated segmentation on the doses to OARs was complicated. We suggest that the auto-segmented contours of tissues in close proximity to the target volume need to be carefully checked and corrected when necessary.

Automatic contouring QA method using a deep learning-based autocontouring system

Purpose

To determine the most accurate similarity metric when using an independent system to verify automatically generated contours.

Methods

A reference autocontouring system (primary system to create clinical contours) and a verification autocontouring system (secondary system to test the primary contours) were used to generate a pair of 6 female pelvic structures (UteroCervix [uterus + cervix], CTVn [nodal clinical target volume (CTV)], PAN [para‐aortic lymph nodes], bladder, rectum, and kidneys) on 49 CT scans from our institution and 38 from other institutions. Additionally, clinically acceptable and unacceptable contours were manually generated using the 49 internal CT scans. Eleven similarity metrics (volumetric Dice similarity coefficient (DSC), Hausdorff distance, 95% Hausdorff distance, mean surface distance, and surface DSC with tolerances from 1 to 10 mm) were calculated between the reference and the verification autocontours, and between the manually generated and the verification autocontours. A support vector machine (SVM) was used to determine the threshold that separates clinically acceptable and unacceptable contours for each structure. The 11 metrics were investigated individually and in certain combinations. Linear, radial basis function, sigmoid, and polynomial kernels were tested using the combinations of metrics as inputs for the SVM.

Results

The highest contouring error detection accuracies were 0.91 for the UteroCervix, 0.90 for the CTVn, 0.89 for the PAN, 0.92 for the bladder, 0.95 for the rectum, and 0.97 for the kidneys and were achieved using surface DSCs with a thickness of 1, 2, or 3 mm. The linear kernel was the most accurate and consistent when a combination of metrics was used as an input for the SVM. However, the best model accuracy from the combinations of metrics was not better than the best model accuracy from a surface DSC as an input.

Conclusions

We distinguished clinically acceptable contours from clinically unacceptable contours with an accuracy higher than 0.9 for the targets and critical structures in patients with cervical cancer; the most accurate similarity metric was surface DSC with a thickness of 1, 2, or 3 mm.

Evaluating the clinical acceptability of deep learning contours of prostate and organs-at-risk in an automated prostate treatment planning process

BACKGROUND

Radiation treatment is considered an effective and the most common treatment option for prostate cancer. The treatment planning process requires accurate and precise segmentation of the prostate and organs at risk (OARs), which is laborious and time-consuming when contoured manually. Artificial intelligence (AI)-based auto-segmentation has the potential to significantly accelerate the radiation therapy treatment planning process; however, the accuracy of auto-segmentation needs to be validated before its full clinical adoption.

PURPOSE

A commercial AI-based contouring model was trained to provide segmentation of the prostate and surrounding OARs. The segmented structures were input to a commercial auto-planning module for automated prostate treatment planning. This study comprehensively evaluates the performance of this contouring model in the automated prostate treatment planning process.

METHODS AND MATERIALS

A 3D U-Net-based model (INTContour, Carina AI) was trained and validated on 84 computed tomography (CT) scans and tested on an additional 23 CT scans from patients treated in our local institution. Prostate and OARs contours generated by the AI model (AI contour) were geometrically evaluated against Reference contours. The prostate contours were further evaluated against AI, Reference, and two additional observer contours for comparison using inter-observer variation (IOV) and 3D boundaries discrepancy analyses. A blinded evaluation was introduced to assess subjectively the clinical acceptability of the AI contours. Finally, treatment plans were created from an automated prostate planning workflow using the AI contours and were evaluated for their clinical acceptability following the RTOG-0815 protocol.

RESULTS

The AI contours demonstrated good geometric accuracy on OARs and prostate contours, with average Dice similarity coefficients (DSC) for bladder, rectum, femoral heads, seminal vesicles, and penile bulb of 0.93, 0.85, 0.96, 0.72, and 0.53, respectively. The DSC, 95% directed Hausdorff Distance (HD95), and Mean Surface Distance (MSD) for the prostate were 0.83±0.05, 6.07±1.87 mm, and 2.07±0.73 mm, respectively. No significant differences were found when comparing with IOV. In the double-blinded evaluation, 95.7% of the AI contours were scored as either "Perfect" (34.8%) or "Acceptable" (60.9%), while only one case (4.3%) was scored as "Unacceptable with minor changes required". In total, 69.6% of the AI contours were considered equal to or better than the Reference contours by an independent radiation oncologist. Automated treatment plans created from the AI contours produced similar and clinically-acceptable dosimetric distributions as those from plans created from Reference contours.

CONCLUSIONS

The investigated AI-based commercial model for prostate segmentation demonstrated good performance in clinical practice. Using this model, the implementation of an automated prostate treatment planning process is clinically feasible. This article is protected by copyright. All rights reserved.

Automatic brain segmentation in preterm infants with post-hemorrhagic hydrocephalus using 3D Bayesian U-Net

Post‐hemorrhagic hydrocephalus (PHH) is a severe complication of intraventricular hemorrhage (IVH) in very preterm infants. PHH monitoring and treatment decisions rely heavily on manual and subjective two‐dimensional measurements of the ventricles. Automatic and reliable three‐dimensional (3D) measurements of the ventricles may provide a more accurate assessment of PHH, and lead to improved monitoring and treatment decisions. To accurately and efficiently obtain these 3D measurements, automatic segmentation of the ventricles can be explored. However, this segmentation is challenging due to the large ventricular anatomical shape variability in preterm infants diagnosed with PHH. This study aims to (a) propose a Bayesian U‐Net method using 3D spatial concrete dropout for automatic brain segmentation (with uncertainty assessment) of preterm infants with PHH; and (b) compare the Bayesian method to three reference methods: DenseNet, U‐Net, and ensemble learning using DenseNets and U‐Nets. A total of 41 T₂‐weighted MRIs from 27 preterm infants were manually segmented into lateral ventricles, external CSF, white and cortical gray matter, brainstem, and cerebellum. These segmentations were used as ground truth for model evaluation. All methods were trained and evaluated using 4‐fold cross‐validation and segmentation endpoints, with additional uncertainty endpoints for the Bayesian method. In the lateral ventricles, segmentation endpoint values for the DenseNet, U‐Net, ensemble learning, and Bayesian U‐Net methods were mean Dice score = 0.814 ± 0.213, 0.944 ± 0.041, 0.942 ± 0.042, and 0.948 ± 0.034 respectively. Uncertainty endpoint values for the Bayesian U‐Net were mean recall = 0.953 ± 0.037, mean  negative predictive value = 0.998 ± 0.005, mean accuracy = 0.906 ± 0.032, and mean AUC = 0.949 ± 0.031. To conclude, the Bayesian U‐Net showed the best segmentation results across all methods and provided accurate uncertainty maps. This method may be used in clinical practice for automatic brain segmentation of preterm infants with PHH, and lead to better PHH monitoring and more informed treatment decisions.

Machine-assisted interpolation algorithm for semi-automated segmentation of highly deformable organs

PURPOSE

Accurate and robust auto-segmentation of highly deformable organs (HDOs), for example, stomach or bowel, remains an outstanding problem due to these organs' frequent and large anatomical variations. Yet, time-consuming manual segmentation of these organs presents a particular challenge to time-limited modern radiotherapy techniques such as on-line adaptive radiotherapy and high-dose-rate brachytherapy. We propose a machine-assisted interpolation (MAI) that uses prior information in the form of sparse manual delineations to facilitate rapid, accurate segmentation of the stomach from low field magnetic resonance images (MRI) and the bowel from computed tomography (CT) images.

METHODS

Stomach MR images from 116 patients undergoing 0.35T MRI-guided abdominal radiotherapy and bowel CT images from 120 patients undergoing high dose rate pelvic brachytherapy treatment were collected. For each patient volume, the manual delineation of the HDO was extracted from every 8th slice. These manually drawn contours were first interpolated to obtain an initial estimate of the HDO contour. A two-channel 64 × 64 pixel patch-based convolutional neural network (CNN) was trained to localize the position of the organ's boundary on each slice within a five-pixel wide road using the image and interpolated contour estimate. This boundary prediction was then input, in conjunction with the image, to an organ closing CNN which output the final organ segmentation. A Dense-UNet architecture was used for both networks. The MAI algorithm was separately trained for the stomach segmentation and the bowel segmentation. Algorithm performance was compared against linear interpolation (LI) alone and against fully automated segmentation (FAS) using a Dense-UNet trained on the same datasets. The Dice Similarity Coefficient (DSC) and mean surface distance (MSD) metrics were used to compare the predictions from the three methods. Statistically significance was tested using Student's t test.

RESULTS

For the stomach segmentation, the mean DSC from MAI (0.91 ± 0.02) was 5.0% and 10.0% higher as compared to LI and FAS, respectively. The average MSD from MAI (0.77 ± 0.25 mm) was 0.54 and 3.19 mm lower compared to the two other methods. Only 7% of MAI stomach predictions resulted in a DSC < 0.8, as compared to 30% and 28% for LI and FAS, respectively. For the bowel segmentation, the mean DSC of MAI (0.90 ± 0.04) was 6% and 18% higher, and the average MSD of MAI (0.93 ± 0.48 mm) was 0.42 and 4.9 mm lower as compared to LI and FAS. Sixteen percent of the predicted contour from MAI resulted in a DSC < 0.8, as compared to 46% and 60% for FAS and LI, respectively. All comparisons between MAI and the baseline methods were found to be statistically significant (p-value < 0.001).

CONCLUSIONS

The proposed MAI algorithm significantly outperformed LI in terms of accuracy and robustness for both stomach segmentation from low-field MRIs and bowel segmentation from CT images. At this time, FAS methods for HDOs still require significant manual editing. Therefore, we believe that the MAI algorithm has the potential to expedite the process of HDO delineation within the radiation therapy workflow.

Assessment of MRI-Linac Economics under the RO-APM

The implementation of the radiation oncology alternative payment model (RO-APM) has raised concerns regarding the development of MRI-guided adaptive radiotherapy (MRgART). We sought to compare technical fee reimbursement under Fee-For-Service (FFS) to the proposed RO-APM for a typical MRI-Linac (MRL) patient load and distribution of 200 patients. In an exploratory aim, a modifier was added to the RO-APM (mRO-APM) to account for the resources necessary to provide this care. Traditional Medicare FFS reimbursement rates were compared to the diagnosis-based reimbursement in the RO-APM. Reimbursement for all selected diagnoses were lower in the RO-APM compared to FFS, with the largest differences in the adaptive treatments for lung cancer (−89%) and pancreatic cancer (−83%). The total annual reimbursement discrepancy amounted to −78%. Without implementation of adaptive replanning there was no difference in reimbursement in breast, colorectal and prostate cancer between RO-APM and mRO-APM. Accommodating online adaptive treatments in the mRO-APM would result in a reimbursement difference from the FFS model of −47% for lung cancer and −46% for pancreatic cancer, mitigating the overall annual reimbursement difference to −54%. Even with adjustment, the implementation of MRgART as a new treatment strategy is susceptible under the RO-APM.

Capturing patient anatomy for designing and manufacturing personalized prostheses

Prostheses play a critical role in healthcare provision for many patients and encompass aesthetic facial prostheses, prosthetic limbs and prosthetic joints, bones, and other implantable medical devices in musculoskeletal surgery. An increasingly important component in cutting-edge healthcare treatments is the ability to accurately capture patient anatomy in order to guide the manufacture of personalized prostheses. This article examines methods for capturing patient anatomy and discusses the degrees of personalization in medical manufacturing alongside a summary of current trends in scanning technology with a focus on identifying workflows for incorporating personalization into patient-specific products. Over the next decade, with increased harmonization of both personalization and automated prosthetic manufacturing will be the realization of improved patient compliance, satisfaction, and clinical outcomes.

Adaptive Radiotherapy in the Management of Cervical Cancer: Review of Strategies and Clinical Implementation

The complex and varied motion of the cervix-uterus target during external beam radiotherapy (EBRT) underscores the clinical benefits afforded by adaptive radiotherapy (ART) techniques. These gains have already been realised in the implementation of image-guided adaptive brachytherapy, where adapting to anatomy at each fraction has seen improvements in clinical outcomes and a reduction in treatment toxicity. With regards to EBRT, multiple adaptive strategies have been implemented, including a personalised internal target volume, offline replanning and a plan of the day approach. With technological advances, there is now the ability for real-time online ART using both magnetic resonance imaging and computed tomography-guided imaging. However, multiple challenges remain in the widespread dissemination of ART. This review investigates the ART strategies and their clinical implementation in EBRT delivery for cervical cancer.

A review of medical image data augmentation techniques for deep learning applications

Research in artificial intelligence for radiology and radiotherapy has recently become increasingly reliant on the use of deep learning-based algorithms. While the performance of the models which these algorithms produce can significantly outperform more traditional machine learning methods, they do rely on larger datasets being available for training. To address this issue, data augmentation has become a popular method for increasing the size of a training dataset, particularly in fields where large datasets aren't typically available, which is often the case when working with medical images. Data augmentation aims to generate additional data which is used to train the model and has been shown to improve performance when validated on a separate unseen dataset. This approach has become commonplace so to help understand the types of data augmentation techniques used in state-of-the-art deep learning models, we conducted a systematic review of the literature where data augmentation was utilised on medical images (limited to CT and MRI) to train a deep learning model. Articles were categorised into basic, deformable, deep learning or other data augmentation techniques. As artificial intelligence models trained using augmented data make their way into the clinic, this review aims to give an insight to these techniques and confidence in the validity of the models produced.

Simple Python Module for Conversions Between DICOM Images and Radiation Therapy Structures, Masks, and Prediction Arrays

Deep learning is becoming increasingly popular and available to new users, particularly in the medical field. Deep learning image segmentation, outcome analysis, and generators rely on presentation of Digital Imaging and Communications in Medicine (DICOM) images and often radiation therapy (RT) structures as masks. Although the technology to convert DICOM images and RT structures into other data types exists, no purpose-built Python module for converting NumPy arrays into RT structures exists. The 2 most popular deep learning libraries, Tensorflow and PyTorch, are both implemented within Python, and we believe a set of tools built in Python for manipulating DICOM images and RT structures would be useful and could save medical researchers large amounts of time and effort during the preprocessing and prediction steps. Our module provides intuitive methods for rapid data curation of RT-structure files by identifying unique region of interest (ROI) names and ROI structure locations and allowing multiple ROI names to represent the same structure. It is also capable of converting DICOM images and RT structures into NumPy arrays and SimpleITK Images, the most commonly used formats for image analysis and inputs into deep learning architectures and radiomic feature calculations. Furthermore, the tool provides a simple method for creating a DICOM RT-structure from predicted NumPy arrays, which are commonly the output of semantic segmentation deep learning models. Accessing DicomRTTool via the public Github project invites open collaboration, and the deployment of our module in PyPi ensures painless distribution and installation. We believe our tool will be increasingly useful as deep learning in medicine progresses.