Biology subjects

Thor, M.

Publications and source records attributed to Thor, M..

3 recordsLinked to original sources

bioRxiv2019

Library of model implementations for sharing deep-learning image segmentation and outcomes models

An open-source library of implementations for deep-learning based image segmentation and outcomes models is presented in this work. As oncology treatment planning becomes increasingly driven by automation, such a library of model implementations is crucial to (i) validate existing models on datasets collected at different institutions, (ii) automate segmentation, (iii) create ensembles for improving performance and (iv) incorporate validated models in the clinical workflow. The library was developed with Computational Environment for Radiological Research (CERR) software platform. CERR is a natural choice to centralize model implementations due to its comprehensiveness, popularity, and ease of use. CERR provides well-validated feature extraction for radiotherapy dosimetry and radiomics with fine control over the calculation settings. This allows users to select the appropriate feature calculation used in the model derivation. Models for automatic image segmentation are distributed via Singularity containers, with seamless i/o to and from CERR. Singularity containers allow for segmentation models to be deployed with a variety of scientific computing architectures. Deployment of models is driven by JSON configuration file, making it convenient to plug-in models. Models from the library can be called programmatically for batch evaluation. The library includes implementations for popular radiotherapy models outlined in the Quantitative Analysis of Normal Tissue Effects in the Clinic effort and recently published literature. Radiomics models include features from Image Biomarker Standardization features found to be important across multiple sites and image modalities. Deep learning-based image segmentation models include state of the art networks such as Deeplab and other problem-specific architectures. The library is distributed as GNU-copyrighted software at https://www.github.com/cerr/CERR.

bioinformatics↗

bioRxiv2019

Deep learning-based auto-segmentation of swallowing and chewing structures in CT

PurposeDelineating the swallowing and chewing structures in Head and Neck (H&N) CT scans is necessary for radiotherapy treatment (RT) planning to reduce the incidence of radiation-induced dysphagia, trismus, and speech dysfunction. Automating this process would decrease the manual input required and yield reproducible segmentations, but generating accurate segmentations is challenging due to the complex morphology of swallowing and chewing structures and limited soft tissue contrast in CT images. MethodsWe trained deep learning models using 194 H&N CT scans from our institution to segment the masseters (left and right), medial pterygoids (left and right), larynx, and pharyngeal constrictor muscle using DeepLabV3+ with the resnet-101 backbone. Models were trained in a sequential manner to guide the localization of each structure group based on prior segmentations. Additionally, an ensemble of models was developed using contextual information from three different views (axial, coronal, and sagittal), for robustness to occasional failures of the individual models. Output probability maps were averaged, and voxels were assigned labels corresponding to the class with the highest combined probability. ResultsThe median dice similarity coefficients (DSC) computed on a hold-out set of 24 CT scans were 0.87{+/-}0.02 for the masseters, 0.80{+/-}0.03 for the medial pterygoids, 0.81{+/-}0.04 for the larynx, and 0.69{+/-}0.07for the constrictor muscle. The corresponding 95th percentile Hausdorff distances were 0.32{+/-}0.08cm (masseters), 0.42{+/-}0.2cm (medial pterygoids), 0.53{+/-}0.3cm (larynx), and 0.36{+/-}0.15cm (constrictor muscle). Dose-volume histogram (DVH) metrics previously found to correlate with each toxicity were extracted from manual and auto-generated contours and compared between the two sets of contours to assess clinical utility. Differences in DVH metrics were not found to be statistically significant (p>0.05) for any of the structures. Further, inter-observer variability in contouring was studied in 10 CT scans. Automated segmentations were found to agree better with each of the observers as compared to inter-observer agreement, measured in terms of DSC. ConclusionsWe developed deep learning-based auto-segmentation models for swallowing and chewing structures in CT. The resulting segmentations can be included in treatment planning to limit complications following RT for H&N cancer. The segmentation models developed in this work are distributed for research use through the open-source platform CERR, accessible at https://github.com/cerr/CERR.

bioinformatics↗

bioRxiv2019

Serum Alpha-2-Macroglobulin as an intrinsic radioprotective factor in patients undergoing thoracic radiation therapy

ObjectiveTo investigate the impact of alpha-2-macroglobulin (A2M), a suspected intrinsic radioprotectant, on radiation pneumonitis and esophagitis. Additionally, we establish multifactorial predictive models for pneumonitis and esophagitis.\n\nMaterials/MethodsBaseline A2M levels were obtained for 258 patients prior to thoracic radiotherapy (RT). Dose-volume characteristics were extracted from treatment plans. Spearmans correlation (Rs) test was used to correlate A2M levels, smoking status and dosimetric variables with toxicities. Esophagitis and pneumonitis prediction models were built using least absolute shrinkage and selection operator (LASSO) logistic regression on 1000 bootstrapped datasets. Models were built using 2/3 of the data for training and 1/3 for validation.\n\nResultsThere were 36 (14.0%) patients with grade [≥]2 pneumonitis and 61 (23.6%) with grade [≥]2 esophagitis. The median A2M level was 191 mg/dL (range: 94-511). Never/former/current smoker status was 47 (18.2%)/179 (69.4%)/32 (12.4%). We found a significant correlation between baseline A2M levels and esophagitis (Rs=-0.18/p=0.003) and between A2M and smoking status (former or current) (Rs=0.13/p=0.04) but not between A2M and pneumonitis. On univariate analysis, significant parameters for grade [≥]2 esophagitis included number of fractions (Rs=0.47/p<0.0001), treatment days (Rs=0.44/p<0.0001), chemotherapy use (Rs=0.40/p<0.0001), dose per fraction (Rs=-0.34/p<0.0001), total dose (Rs=0.29/p<0.0001), age (Rs=-0.22/p=0.0003), and several dosimetric variables in esophagus with Rs>0.5 (p<0.0001). For pneumonitis, significant clinical parameters were treatment days (Rs=0.24/p=0.0001), chemotherapy use (Rs=0.22/p=0.0004), number of fractions (Rs=0.21/p=0.0007), dose per fraction (Rs=-0.18/p=0.0035), and total dose (Rs=0.15/p=0.013). The most significant dosimetric variable in lung and heart was D70 (Rs=0.28/p<0.0001) and max dose (Rs=0.27/p<0.0001), respectively. LASSO bootstrap logistic regression models on the validation data resulted in the area under the receiver operating characteristic curve of 0.84 and 0.75 for esophagitis and pneumonitis, respectively.\n\nConclusionOur findings show an association of higher A2M values with lower risk of radiation esophagitis and smoking status. Multivariate predictive models also confirmed a role of heart dose in the risk of pneumonitis.

cancer biology↗