Search

The Dice Similarity Coefficient (DSC) is the current de facto standard to determine agreement between a reference segmentation and one generated by manual / auto-contouring approaches. This metric is useful for non-spatially important images; however, radiation therapy requires consideration of nearby Organs-at-Risk (OARs) and their radiosensitivity which are currently unaccounted for with the traditional DSC. In this work, we introduce the OAR-DSC which accounts for nearby OARs and their radiosensitivity when computing the DSC. We illustrate the importance of this through cases where two proposed contours have similar DSC, but lower OAR-DSC when one contour expands closer to the surrounding OARs. This work is important because the OAR-DSC may be used by deep learning auto-contouring algorithms in a radiation therapy specific loss function, thereby progressing on the current disregard for the importance of these differences on the final radiation dose plan generation, delivery, and risks of patient toxicity., Comment: Proceedings of the XX-th international conference on the use of computers in radiation therapy (ICCR)

The main treatment modality for oropharyngeal cancer (OPC) is radiotherapy, where accurate segmentation of the primary gross tumor volume (GTVp) is essential. However, accurate GTVp segmentation is challenging due to significant interobserver variability and the time-consuming nature of manual annotation, while fully automated methods can occasionally fail. An interactive deep learning (DL) model offers the advantage of automatic high-performance segmentation with the flexibility for user correction when necessary. In this study, we examine interactive DL for GTVp segmentation in OPC. We implement state-of-the-art algorithms and propose a novel two-stage Interactive Click Refinement (2S-ICR) framework. Using the 2021 HEad and neCK TumOR (HECKTOR) dataset for development and an external dataset from The University of Texas MD Anderson Cancer Center for evaluation, the 2S-ICR framework achieves a Dice similarity coefficient of 0.713 $\pm$ 0.152 without user interaction and 0.824 $\pm$ 0.099 after five interactions, outperforming existing methods in both cases.

Digital twin models are of high interest to Head and Neck Cancer (HNC) oncologists, who have to navigate a series of complex treatment decisions that weigh the efficacy of tumor control against toxicity and mortality risks. Evaluating individual risk profiles necessitates a deeper understanding of the interplay between different factors such as patient health, spatial tumor location and spread, and risk of subsequent toxicities that can not be adequately captured through simple heuristics. To support clinicians in better understanding tradeoffs when deciding on treatment courses, we developed DITTO, a digital-twin and visual computing system that allows clinicians to analyze detailed risk profiles for each patient, and decide on a treatment plan. DITTO relies on a sequential Deep Reinforcement Learning digital twin (DT) to deliver personalized risk of both long-term and short-term disease outcome and toxicity risk for HNC patients. Based on a participatory collaborative design alongside oncologists, we also implement several visual explainability methods to promote clinical trust and encourage healthy skepticism when using our system. We evaluate the efficacy of DITTO through quantitative evaluation of performance and case studies with qualitative feedback. Finally, we discuss design lessons for developing clinical visual XAI applications for clinical end users.

Deep learning has significantly advanced the potential for automated contouring in radiotherapy planning. In this manuscript, guided by contemporary literature, we underscore three key insights: (1) High-quality training data is essential for auto-contouring algorithms; (2) Auto-contouring models demonstrate commendable performance even with limited medical image data; (3) The quantitative performance of auto-contouring is reaching a plateau. Given these insights, we emphasize the need for the radiotherapy research community to embrace data-centric approaches to further foster clinical adoption of auto-contouring technologies., Comment: 13 pages, 4 figures

Developing applicable clinical machine learning models is a difficult task when the data includes spatial information, for example, radiation dose distributions across adjacent organs at risk. We describe the co-design of a modeling system, DASS, to support the hybrid human-machine development and validation of predictive models for estimating long-term toxicities related to radiotherapy doses in head and neck cancer patients. Developed in collaboration with domain experts in oncology and data mining, DASS incorporates human-in-the-loop visual steering, spatial data, and explainable AI to augment domain knowledge with automatic data mining. We demonstrate DASS with the development of two practical clinical stratification models and report feedback from domain experts. Finally, we describe the design lessons learned from this collaborative experience., Comment: 10 pages, 9 figures

International benchmarking competitions have become fundamental for the comparative performance assessment of image analysis methods. However, little attention has been given to investigating what can be learnt from these competitions. Do they really generate scientific progress? What are common and successful participation strategies? What makes a solution superior to a competing method? To address this gap in the literature, we performed a multi-center study with all 80 competitions that were conducted in the scope of IEEE ISBI 2021 and MICCAI 2021. Statistical analyses performed based on comprehensive descriptions of the submitted algorithms linked to their rank as well as the underlying participation strategies revealed common characteristics of winning solutions. These typically include the use of multi-task learning (63%) and/or multi-stage pipelines (61%), and a focus on augmentation (100%), image preprocessing (97%), data curation (79%), and postprocessing (66%). The "typical" lead of a winning team is a computer scientist with a doctoral degree, five years of experience in biomedical image analysis, and four years of experience in deep learning. Two core general development strategies stood out for highly-ranked teams: the reflection of the metrics in the method design and the focus on analyzing and handling failure cases. According to the organizers, 43% of the winning algorithms exceeded the state of the art but only 11% completely solved the respective domain problem. The insights of our study could help researchers (1) improve algorithm development strategies when approaching new problems, and (2) focus on open research questions revealed by this work., Comment: accepted to CVPR 2023

Purpose: This study aimed to use deep learning-based dose prediction to assess head and neck (HN) plan quality and identify suboptimal plans. Methods: A total of 245 VMAT HN plans were created using RapidPlan knowledge-based planning (KBP). A subset of 112 high-quality plans was selected under the supervision of an HN radiation oncologist. We trained a 3D Dense Dilated U-Net architecture to predict 3-dimensional dose distributions using 3-fold cross-validation on 90 plans. Model inputs included CT images, target prescriptions, and contours for targets and organs at risk (OARs). The model's performance was assessed on the remaining 22 test plans. We then tested the application of the dose prediction model for automated review of plan quality. Dose distributions were predicted on 14 clinical plans. The predicted versus clinical OAR dose metrics were compared to flag OARs with suboptimal normal tissue sparing using a 2 Gy dose difference or 3% dose-volume threshold. OAR flags were compared to manual flags by 3 HN radiation oncologists. Results: The predicted dose distributions were of comparable quality to the KBP plans. The differences between the predicted and KBP-planned D1%, D95%, and D99% across the targets were within -2.53%(SD=1.34%), -0.42%(SD=1.27%), and -0.12%(SD=1.97%), respectively, and the OAR mean and maximum doses were within -0.33Gy(SD=1.40Gy) and -0.96Gy(SD=2.08Gy). For the plan quality assessment study, radiation oncologists flagged 47 OARs for possible plan improvement. There was high interphysician variability; 83% of physician-flagged OARs were flagged by only one of 3 physicians. The comparative dose prediction model flagged 63 OARs, including 30 of 47 physician-flagged OARs. Conclusion: Deep learning can predict high-quality dose distributions, which can be used as comparative dose distributions for automated, individualized assessment of HN plan quality., Comment: updated to reflect the published peer-reviewed article

Chemotherapy is one of the primary modalities of cancer treatment. Chemotherapy drug administration is a complex problem that often requires expensive clinical trials to evaluate potential regimens. One way to alleviate this burden and better inform future trials is to build reliable models for drug administration. Previous chemotherapy optimization models have mainly relied on optimal control, which does not lend itself to capturing complex and vital operational constraints in chemotherapy planning involving discrete decisions, such as doses via pills and rest periods. In addition, most of the existing models for chemotherapy optimization lack an explicit toxicity measure and impose toxicity constraints primarily through (fixed) limits on drug concentration. The existing stochastic optimization models also focus on maximizing the probability of cure when tumor heterogeneity is uncertain. In this paper, we develop a mixed-integer program for combination chemotherapy (utilization of multiple drugs) optimization that incorporates various important operational constraints and, besides dose and concentration limits, controls treatment toxicity based on its effect on the count of white blood cells. To address the uncertainty of tumor heterogeneity, we propose chance constraints that guarantee reaching an operable tumor size with a high probability in a neoadjuvant setting. We present analytical results pertinent to the accuracy of the model in representing biological processes of chemotherapy and establish its merit for clinical applications through a numerical study of breast cancer.

Advances in data collection in radiation therapy have led to an abundance of opportunities for applying data mining and machine learning techniques to promote new data-driven insights. In light of these advances, supporting collaboration between machine learning experts and clinicians is important for facilitating better development and adoption of these models. Although many medical use-cases rely on spatial data, where understanding and visualizing the underlying structure of the data is important, little is known about the interpretability of spatial clustering results by clinical audiences. In this work, we reflect on the design of visualizations for explaining novel approaches to clustering complex anatomical data from head and neck cancer patients. These visualizations were developed, through participatory design, for clinical audiences during a multi-year collaboration with radiation oncologists and statisticians. We distill this collaboration into a set of lessons learned for creating visual and explainable spatial clustering for clinical users.

MRI offers outstanding soft tissue contrast that may reduce uncertainties in target and organ-at-risk delineation and enable online adaptive image-guided treatment. Spatial distortions resulting from non-linearities in the gradient fields and non-uniformity in the main magnetic field must be accounted for across the imaging field-of-view to prevent systematic errors during treatment delivery. This work presents a modular phantom and software application to characterize geometric distortion (GD) within the large field-of-view MRI images required for radiation therapy simulation. The modular phantom is assembled from a series of rectangular foam blocks containing high-contrast fiducial markers in a known configuration. The modular phantom design facilitates transportation of the phantom between different MR scanners and MR-guided linear accelerators and allows the phantom to be adapted to fit different sized bores or coils. The phantom was evaluated using a 1.5T MR-guided linear accelerator (MR-Linac) and 1.5T and 3.0T diagnostic scanners. Performance was assessed by varying acquisition parameters to induce image distortions in a known manner. Imaging was performed using T1 and T2 weighted pulse sequences with 2D and 3D distortion correction algorithms and the receiver bandwidth (BW) varied as 250-815 Hz/pixel. Phantom set-up reproducibility was evaluated across independent set-ups. The software was validated by comparison with a non-modular phantom. Average geometric distortion was 0.94+/-0.58 mm for the MR-Linac, 0.90+/-0.53 mm for the 1.5 T scanner, and 1.15+/-0.62 mm for the 3.0T scanner, for a 400 mm diameter volume-of-interest. GD increased, as expected, with decreasing BW, and with the 2D versus 3D correction algorithm. Differences in GD attributed to phantom set-up were 0.13 mm or less. Differences in GD for the two software applications were less than 0.07 mm., Comment: 25 pages

In the context of a binary classification problem, the optimal linear combination of continuous predictors can be estimated by maximizing an empirical estimate of the area under the receiver operating characteristic (ROC) curve (AUC). For multi-category responses, the optimal predictor combination can similarly be obtained by maximization of the empirical hypervolume under the manifold (HUM). This problem is particularly relevant to medical research, where it may be of interest to diagnose a disease with various subtypes or predict a multi-category outcome. Since the empirical HUM is discontinuous, non-differentiable, and possibly multi-modal, solving this maximization problem requires a global optimization technique. Estimation of the optimal coefficient vector using existing global optimization techniques is computationally expensive, becoming prohibitive as the number of predictors and the number of outcome categories increases. We propose an efficient derivative-free black-box optimization technique based on pattern search to solve this problem. Through extensive simulation studies, we demonstrate that the proposed method achieves better performance compared to existing methods including the step-down algorithm. Finally, we illustrate the proposed method to predict swallowing difficulty after radiation therapy for oropharyngeal cancer based on radiation dose to various structures in the head and neck.