Publications by authors named "Raymond P Mumme"
Purpose: To evaluate the efficacy of prominent machine learning algorithms in predicting normal tissue complication probability using clinical data obtained from 2 distinct disease sites and to create a software tool that facilitates the automatic determination of the optimal algorithm to model any given labeled data set.
Methods And Materials: We obtained 3 sets of radiation toxicity data (478 patients) from our clinic: gastrointestinal toxicity, radiation pneumonitis, and radiation esophagitis. These data comprised clinicopathological and dosimetric information for patients diagnosed with non-small cell lung cancer and anal squamous cell carcinoma.
Article Synopsis
- Radiation treatment planning is complicated and can vary significantly between different planners, but knowledge-based planning (KBP) aims to streamline the process and produce high-quality plans regardless of the planner's skills.
- The study involved creating and validating 10 automated KBP models for various treatment sites, which incorporated advanced planning scripts and optimization techniques to operate without human input.
- The results showed that 88% of the automated plans were deemed "acceptable as is" by physicians, indicating that this approach could significantly improve the efficiency and consistency of radiation treatment planning.
Article Synopsis
- The study aimed to create a deep learning model for accurately segmenting cardiac substructures to reduce radiation risks in lung cancer treatment.
- Manual delineation of nineteen cardiac substructures from 100 patients was performed and compared to the results from the trained nnU-Net model using various evaluation metrics.
- The model showed high accuracy with an average Dice similarity coefficient of 0.95 for the whole heart and 94% of the auto-segmented contours were deemed clinically acceptable by physicians, suggesting it’s a promising tool for future research on radiation dose effects.
In this study, we aimed to enhance the contouring accuracy of cardiac pacemakers by improving their visualization using deep learning models to predict MV CBCT images based on kV CT or CBCT images. Ten pacemakers and four thorax phantoms were included, creating a total of 35 combinations. Each combination was imaged on a Varian Halcyon (kV/MV CBCT images) and Siemens SOMATOM CT scanner (kV CT images).
View Article and Find Full Text PDFPurpose: Variability in contouring structures of interest for radiotherapy continues to be challenging. Although training can reduce such variability, having radiation oncologists provide feedback can be impractical. We developed a contour training tool to provide real-time feedback to trainees, thereby reducing variability in contouring.
View Article and Find Full Text PDFPurpose: Our purpose was to analyze the effect on gastrointestinal (GI) toxicity models when their dose-volume metrics predictors are derived from segmentations of the peritoneal cavity after different contouring approaches.
Methods And Materials: A random forest machine learning approach was used to predict acute grade ≥3 GI toxicity from dose-volume metrics and clinicopathologic factors for 246 patients (toxicity incidence = 9.5%) treated with definitive chemoradiation for squamous cell carcinoma of the anus.
Manually delineating upper abdominal organs at risk (OARs) is a time-consuming task. To develop a deep-learning-based tool for accurate and robust auto-segmentation of these OARs, forty pancreatic cancer patients with contrast-enhanced breath-hold computed tomographic (CT) images were selected. We trained a three-dimensional (3D) U-Net ensemble that automatically segments all organ contours concurrently with the self-configuring nnU-Net framework.
View Article and Find Full Text PDFArticle Synopsis
- Vertebral labelling and segmentation are crucial for improving automated spine image processing, aiding in clinical decision-making and population health analysis.
- The Large Scale Vertebrae Segmentation Challenge (VerSe) was created to tackle the challenges of this field by having participants develop algorithms for labelling and segmenting vertebrae using a curated dataset of CT scans.
- Results showed that an algorithm's performance depends significantly on its ability to identify vertebrae with rare anatomical variations, highlighting the complexities in spine analysis.