{"id":{"repo_id":"buffalo","oai_identifier":"oai:ubir.buffalo.edu:10477/86680"},"canonical_url":"https://search.dev.ndltd.org/etd/buffalo/oai:ubir.buffalo.edu:10477/86680","repository":{"repo_id":"buffalo","name":"Buffalo","base_url":"https://ubir.buffalo.edu/oai/request"},"display":{"title":"A Comparative Study of Novel Deep Learning-Based and Conventional Atlas-Based Automatic Segmentation in Head and Neck Radiotherapy Planning","abstract":"M.S.","abstract_html":"M.S.","abstract_has_math":false,"creators":["Asbach, John; 0000-0002-7300-5967"],"institution":"State University of New York at Buffalo","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Le, Anh","Radiology"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-02-21T21:36:23Z","date_published":"2025-02-21T21:36:23Z","updated_at":"2026-07-27T19:05:34Z","subjects":["medical imaging","medical physics","oncology","computer science"],"languages":["eng"],"rights":["Users of works found in University at Buffalo Institutional Repository (UBIR) are responsible for identifying and contacting the copyright owner for permission to reuse. University at Buffalo Libraries do not manage rights for copyright-protected works and cannot assist with permissions.","Copyright retained by author."],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/10477/86680","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Le, Anh","Radiology"]},{"key":"dc:creator","label":"Author","values":["Asbach, John; 0000-0002-7300-5967"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2025-02-21T21:36:23Z","2020"]},{"key":"dc:publisher","label":"Institution","values":["State University of New York at Buffalo"]},{"key":"dc:type","label":"Dc Type","values":["Text","Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["medical imaging","medical physics","oncology","computer science"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Users of works found in University at Buffalo Institutional Repository (UBIR) are responsible for identifying and contacting the copyright owner for permission to reuse. University at Buffalo Libraries do not manage rights for copyright-protected works and cannot assist with permissions.","Copyright retained by author."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["http://hdl.handle.net/10477/86680"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["M.S.","Purpose: To identify opportunities to improve automatic contouring of organs at risk for head and neck cancer patients by developing a convolutional neural network to conduct automatic contouring and comparing its performance to contemporary atlas-based methods. Methods: Using a cohort of 265 anonymized planning CT scans and their associated physician-defined structure set, with 19 set aside for evaluation, a convolutional neural network (CNN) was trained to predict contour data for eleven common organs at risk (OAR) in head and neck cancer patients. Once trained, the neural network predicted contours for the 19 CT scans that were in the evaluation dataset. A commercial deformable atlas-based automatic contour generation function was used to generate competing contour data for the same 19 CT scans. Each output was compared to approved physician-generated manual contours across three quantitative metrics: Dice similarity coefficient (DSC), mean surface distance (MSD), and 95th percentile Hausdorff distance (HD). Performance differences between the modalities on a given metric for each OAR were evaluated for statistical significance using a two-tailed t-test with a p-value threshold of 0.05 to define a statistically significant difference between the modalities. Results: The CNN scored better on all three metrics for both parotid glands and both submandibular glands. For the cochlea and brachial plexus, the CNN scored better on some metrics, and on some metrics, there was no statistically significant difference between the two modalities. For the brain, larynx, and spinal cord, the metrics showed no statistically superior modality. For the brainstem, the CNN contours scored statistically worse than the atlas-based method...Conclusion: Except for the brainstem, the CNN performed either superiorly to or not significantly different from the atlas modality. Deep learning methods demonstrate clear, significant improvement in contouring several OARs. Further pursuit of deep learning techniques for automatic contouring could deliver greater improvements in modern automatic contouring.","**To request an accessible version of the file(s) associated with this item, contact library@buffalo.edu. Please include the item's persistent URL [http://hdl.handle.net/. . .] in your request.**"]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["A Comparative Study of Novel Deep Learning-Based and Conventional Atlas-Based Automatic Segmentation in Head and Neck Radiotherapy Planning"]}]}],"canonical_facts":{"dc:contributor":["Le, Anh","Radiology"],"dc:creator":["Asbach, John; 0000-0002-7300-5967"],"dc:date":["2025-02-21T21:36:23Z","2020"],"dc:description":["M.S.","Purpose: To identify opportunities to improve automatic contouring of organs at risk for head and neck cancer patients by developing a convolutional neural network to conduct automatic contouring and comparing its performance to contemporary atlas-based methods. Methods: Using a cohort of 265 anonymized planning CT scans and their associated physician-defined structure set, with 19 set aside for evaluation, a convolutional neural network (CNN) was trained to predict contour data for eleven common organs at risk (OAR) in head and neck cancer patients. Once trained, the neural network predicted contours for the 19 CT scans that were in the evaluation dataset. A commercial deformable atlas-based automatic contour generation function was used to generate competing contour data for the same 19 CT scans. Each output was compared to approved physician-generated manual contours across three quantitative metrics: Dice similarity coefficient (DSC), mean surface distance (MSD), and 95th percentile Hausdorff distance (HD). Performance differences between the modalities on a given metric for each OAR were evaluated for statistical significance using a two-tailed t-test with a p-value threshold of 0.05 to define a statistically significant difference between the modalities. Results: The CNN scored better on all three metrics for both parotid glands and both submandibular glands. For the cochlea and brachial plexus, the CNN scored better on some metrics, and on some metrics, there was no statistically significant difference between the two modalities. For the brain, larynx, and spinal cord, the metrics showed no statistically superior modality. For the brainstem, the CNN contours scored statistically worse than the atlas-based method...Conclusion: Except for the brainstem, the CNN performed either superiorly to or not significantly different from the atlas modality. Deep learning methods demonstrate clear, significant improvement in contouring several OARs. Further pursuit of deep learning techniques for automatic contouring could deliver greater improvements in modern automatic contouring.","**To request an accessible version of the file(s) associated with this item, contact library@buffalo.edu. Please include the item's persistent URL [http://hdl.handle.net/. . .] in your request.**"],"dc:format":["application/pdf"],"dc:identifier":["http://hdl.handle.net/10477/86680"],"dc:language":["eng"],"dc:publisher":["State University of New York at Buffalo"],"dc:rights":["Users of works found in University at Buffalo Institutional Repository (UBIR) are responsible for identifying and contacting the copyright owner for permission to reuse. University at Buffalo Libraries do not manage rights for copyright-protected works and cannot assist with permissions.","Copyright retained by author."],"dc:subject":["medical imaging","medical physics","oncology","computer science"],"dc:title":["A Comparative Study of Novel Deep Learning-Based and Conventional Atlas-Based Automatic Segmentation in Head and Neck Radiotherapy Planning"],"dc:type":["Text","Thesis"]},"updated_at":"2026-07-27T19:05:34Z"}