Machine learning-based detection of aberrant deep learning segmentations of target and organs at risk for prostate radiotherapy using a secondary segmentation algorithm.

Objective.The output of a deep learning (DL) auto-segmentation application should be reviewed, corrected if needed and approved before being used clinically. This verification procedure is labour-intensive, time-consuming and user-dependent, which potentially leads to significant errors with impact on the overall treatment quality. Additionally, when the time needed to correct auto-segmentations approaches the time to delineate target and organs at risk from scratch, the usability of the DL model can be questioned. Therefore, an automated quality assurance framework was developed with the aim to detect in advance aberrant auto-segmentations.Approach. Five organs (prostate, bladder, anorectum, femoral head left and right) were auto-delineated on CT acquisitions for 48 prostate patients by an in-house trained primary DL model. An experienced radiation oncologist assessed the correctness of the model output and categorised the auto-segmentations into two classes whether minor or major adaptations were needed. Subsequently, an independent, secondary DL model was implemented to delineate the same structures as the primary model. Quantitative comparison metrics were calculated using both models' segmentations and used as input features for a machine learning classification model to predict the output quality of the primary model.Main results. For every organ, the approach of independent validation by the secondary model was able to detect primary auto-segmentations that needed major adaptation with high sensitivity (recall = 1) based on the calculated quantitative metrics. The surface DSC and APL were found to be the most indicated parameters in comparison to standard quantitative metrics for the time needed to adapt auto-segmentations.Significance. This proposed method includes a proof of concept for the use of an independent DL segmentation model in combination with a ML classifier to improve time saving during QA of auto-segmentations. The integration of such system into current automatic segmentation pipelines can increase the efficiency of the radiotherapy contouring workflow.

Towards a safe and efficient clinical implementation of machine learning in radiation oncology by exploring model interpretability, explainability and data-model dependency.

The interest in machine learning (ML) has grown tremendously in recent years, partly due to the performance leap that occurred with new techniques of deep learning, convolutional neural networks for images, increased computational power, and wider availability of large datasets. Most fields of medicine follow that popular trend and, notably, radiation oncology is one of those that are at the forefront, with already a long tradition in using digital images and fully computerized workflows. ML models are driven by data, and in contrast with many statistical or physical models, they can be very large and complex, with countless generic parameters. This inevitably raises two questions, namely, the tight dependence between the models and the datasets that feed them, and the interpretability of the models, which scales with its complexity. Any problems in the data used to train the model will be later reflected in their performance. This, together with the low interpretability of ML models, makes their implementation into the clinical workflow particularly difficult. Building tools for risk assessment and quality assurance of ML models must involve then two main points: interpretability and data-model dependency. After a joint introduction of both radiation oncology and ML, this paper reviews the main risks and current solutions when applying the latter to workflows in the former. Risks associated with data and models, as well as their interaction, are detailed. Next, the core concepts of interpretability, explainability, and data-model dependency are formally defined and illustrated with examples. Afterwards, a broad discussion goes through key applications of ML in workflows of radiation oncology as well as vendors' perspectives for the clinical implementation of ML.

Artificial intelligence and machine learning for medical imaging: A technology review.

Artificial intelligence (AI) has recently become a very popular buzzword, as a consequence of disruptive technical advances and impressive experimental results, notably in the field of image analysis and processing. In medicine, specialties where images are central, like radiology, pathology or oncology, have seized the opportunity and considerable efforts in research and development have been deployed to transfer the potential of AI to clinical applications. With AI becoming a more mainstream tool for typical medical imaging analysis tasks, such as diagnosis, segmentation, or classification, the key for a safe and efficient use of clinical AI applications relies, in part, on informed practitioners. The aim of this review is to present the basic technological pillars of AI, together with the state-of-the-art machine learning methods and their application to medical imaging. In addition, we discuss the new trends and future research directions. This will help the reader to understand how AI methods are now becoming an ubiquitous tool in any medical image analysis workflow and pave the way for the clinical implementation of AI-based solutions.

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

PURPOSE: This study investigated deep learning models for automatic segmentation to support the development of daily online dose optimization strategies, eliminating the need for internal target volume expansions and thereby reducing toxicity events of intensity modulated radiation therapy for cervical cancer. METHODS AND MATERIALS: The cervix-uterus, vagina, parametrium, bladder, rectum, sigmoid, femoral heads, kidneys, spinal cord, and bowel bag were delineated on 408 computed tomography (CT) scans from patients treated at MD Anderson Cancer Center (n = 214), Polyclinique Bordeaux Nord Aquitaine (n = 30), and enrolled in a Medical Image Computing & Computer Assisted Intervention challenge (n = 3). The data were divided into 255 training, 61 validation, 62 internal test, and 30 external test CT scans. Two models were investigated: the 2-dimensional (2D) DeepLabV3+ (Google) and 3-dimensional (3D) Unet in RayStation (RaySearch Laboratories). Three intensity modulated radiation therapy plans were generated on each CT of the internal and external test sets using either the manual, 2D model, or 3D model segmentations. The dose constraints followed the External beam radiochemotherapy and MRI based adaptive BRAchytherapy in locally advanced CErvical cancer (EMBRACE) II protocol, with reduced margins of 5 and 3 mm for the target and nodal planning target volume. Geometric discrepancies between the manual and predicted contours were assessed using the Dice similarity coefficient (DSC), distance-to-agreement, and Hausdorff distance. Dosimetric discrepancies between the manual and model doses were assessed using clinical indices on the manual contours and the gamma index. Interobserver variability was assessed for the cervix-uterus, parametrium, and vagina for the definition of the primary clinical target volume (CTVT) on the external test set. RESULTS: Average DSCs across all organs were 0.67 to 0.96, 0.71 to 0.97, and 0.42 to 0.92 for the 2D model and 0.66 to 0.96, 0.70 to 0.97, and 0.37 to 0.93 for the 3D model on the validation, internal, and external test sets. Average DSCs of the CTVT were 0.88 and 0.81 for the 2D model and 0.87 and 0.82 for the 3D model on the internal and external test sets. Interobserver variability of the CTVT corresponded to a mean (range) DSC of 0.85 (0.77-0.90) on the external test set. On the internal test set, the doses from the 2D and 3D model contours provided a CTVT V42.75 Gy >98% for 98% and 91% of the CT scans, respectively. On the external test set, these percentages were increased to 100% and 93% for the 2D and 3D models, respectively. CONCLUSIONS: The investigated models provided auto-segmentation of the cervix anatomy with similar performances on 2 institutional data sets and reasonable dosimetric accuracies using small planning target volume margins, paving the way to automatic online dose optimization for advanced adaptive radiation therapy strategies.

Stereotactic Ablative Radiation Therapy to All Lesions in Patients With Oligometastatic Cancers: A Phase 1 Dose-Escalation Trial.

PURPOSE: Increasing evidence suggests that patients with a limited number of metastases benefit from SABR to all lesions. However, the optimal dose and fractionation remain unknown. This is particularly true for bone and lymph node metastases. Therefore, a prospective, single-center, dose-escalation trial was initiated. METHODS: Dose-Escalation trial of STereotactic ablative body RadiOtherapY for non-spine bone and lymph node metastases (DESTROY) was an open-label phase 1 trial evaluating SABR to nonspine bone and lymph node lesions in patients with up to 3 metastases. Patients with European Cooperative Oncology Group performance status ≤1, an estimated life expectancy of at least 6 months, and histologically confirmed nonhematological malignancy were eligible. Three SABR fractionation regimens, ie, 5 fractions of 7.0 Gy versus 3 fractions of 10.0 Gy versus a single fraction of 20.0 Gy, were applied in 3 consecutive patient cohorts. The rate of ≥grade 3 toxicity, scored according to the Common Toxicity Criteria for Adverse Events v. 4.03, up to 6 months after SABR, was the primary endpoint. The trial was registered on clinicaltrials.gov (NCT03486431). RESULTS: Between July 2017 and December 2018, 90 patients were enrolled. In total 101 metastases were treated. No ≥grade 3 toxicity was observed in any of the enrolled patients (95% CI 0.0%-12.3% for the first cohort with 28 analyzable patients; 95% CI 0.0%-11.6% for the second and third cohort with 30 analyzable patients each). Treatment-related grade 2 toxicities occurred in 4 out of 30 versus 2 out of 30 versus 2 out of 30 patients for the 5, 3 and 1 fraction schedule, respectively. Actuarial local control rate at 12 months was 94.5%. CONCLUSION: All 3 treatment schedules were feasible and effective with remarkably low toxicity rates and high local control rates. From a patient and resource point of view, the single-fraction schedule is undoubtedly most convenient.

Automated delineation of the clinical target volume using anatomically constrained 3D expansion of the gross tumor volume.

PURPOSE: Delineation of the clinical target volume (CTV) is arguably the weakest link in the treatment planning chain. This work aims to support clinicians in this crucial task. METHODS AND MATERIALS: While the CTV itself is ambiguous, it is much easier to identify structures that do not belong to the CTV and serve as barriers to the spread of the disease. We segment the known barrier structures using a convolutional neural network (CNN). The CTV is then obtained by starting from the manually delineated gross tumor volume (GTV) and expanding it while taking into account the barrier structures. Mathematically, we define the CTV as an iso-surface in the 3D map of shortest paths of all voxels from the GTV. The shortest paths are found with the Dijkstra algorithm. While the method is generally applicable, we test it on 206 glioma and glioblastoma cases. RESULTS: The auto-segmented barrier structures for the brain cases include the ventricles, falx cerebri, tentorium cerebelli, brain sinuses, and the outer surface of the brain. Manual and auto-segmented barrier structures agree with surface Dice Similarity Coefficients (DSC) ranging from 0.91 to 0.97 at 2 mm tolerance. Comparison of manual and automatically delineated CTVs shows a median surface DSC of 0.79. CONCLUSIONS: Barrier structures for CTV definition can be auto-delineated with outstanding precision using a CNN. An algorithm for automated calculation of the CTV by 3D expansion of the GTV while respecting anatomical barriers has been developed. It shows good agreement with manual CTV definition for brain tumors.