{"id":{"repo_id":"utswmed","oai_identifier":"oai:utswmed-ir.tdl.org:2152.5/10586"},"canonical_url":"https://search.dev.ndltd.org/etd/utswmed/oai:utswmed-ir.tdl.org:2152.5/10586","repository":{"repo_id":"utswmed","name":"University of Texas Southwestern Medical Center","base_url":"https://utswmed-ir.tdl.org/server/oai/request"},"display":{"title":"Machine Learning Techniques for Augmenting Hematological Malignancy Diagnosis","abstract":"This dissertation examines the use of modern machine learning (ML) methods to enhance the diagnostic workflow of hematological malignancies, particularly in cytogenetics and clinical flow cytometry. The recent advancements in the field of ML provide a unique opportunity to develop, adapt, and evaluate algorithms for augmenting the hematopathology diagnostic workflow. The field of hematology oncology involves a complex range of diseases, and the final diagnosis is reliant on the rapid analysis of several streams of biological data, including morphologic images, cytogenetic micrographs, molecular sequencing data, and high-dimensional multicolor flow cytometry (MFC). This dissertation explores the various strategies and advantages of using ML to augment this workflow, with a focus on automated screening for clinically informative recurring structural chromosomal abnormalities and serving as a clinical decision support system for clinical MFC data analysis. Most of the data used for this work are from patients with acute myeloid leukemia, but these ML algorithms can easily be extended to other indications. The ML models and techniques applied to cytogenetic micrographs and MFC data are predominantly of the deep learning variety, including convolutional neural networks (CNNs) and graph neural networks (GNNs), along with some other techniques like unsupervised ML, e.g., statistical representation learning and clustering, and graph partitioning. This research provides two key developments for the field. First, that CNNs can accurately detect over a dozen recurrent structural chromosome abnormalities common in several types of acute and chronic leukemia and simultaneously ignore non-recurrent structural abnormalities, laying the groundwork for an autonomous karyotyping system. Second, the first known application a GNN to MFC data is incredibly robust predicting various subtypes of acute promyelocytic leukemia and could be used as part of a clinical decision support system. Importantly, both these studies emphasize generalizability and explainabiliy. These approaches hold promise for facilitating automated analysis in a prospective randomized study and contribute to the digitization and ML revolution taking place in hematopathology clinics worldwide.","abstract_html":"This dissertation examines the use of modern machine learning (ML) methods to enhance the diagnostic workflow of hematological malignancies, particularly in cytogenetics and clinical flow cytometry. The recent advancements in the field of ML provide a unique opportunity to develop, adapt, and evaluate algorithms for augmenting the hematopathology diagnostic workflow. The field of hematology oncology involves a complex range of diseases, and the final diagnosis is reliant on the rapid analysis of several streams of biological data, including morphologic images, cytogenetic micrographs, molecular sequencing data, and high-dimensional multicolor flow cytometry (MFC). This dissertation explores the various strategies and advantages of using ML to augment this workflow, with a focus on automated screening for clinically informative recurring structural chromosomal abnormalities and serving as a clinical decision support system for clinical MFC data analysis. Most of the data used for this work are from patients with acute myeloid leukemia, but these ML algorithms can easily be extended to other indications. The ML models and techniques applied to cytogenetic micrographs and MFC data are predominantly of the deep learning variety, including convolutional neural networks (CNNs) and graph neural networks (GNNs), along with some other techniques like unsupervised ML, e.g., statistical representation learning and clustering, and graph partitioning. This research provides two key developments for the field. First, that CNNs can accurately detect over a dozen recurrent structural chromosome abnormalities common in several types of acute and chronic leukemia and simultaneously ignore non-recurrent structural abnormalities, laying the groundwork for an autonomous karyotyping system. Second, the first known application a GNN to MFC data is incredibly robust predicting various subtypes of acute promyelocytic leukemia and could be used as part of a clinical decision support system. Importantly, both these studies emphasize generalizability and explainabiliy. These approaches hold promise for facilitating automated analysis in a prospective randomized study and contribute to the digitization and ML revolution taking place in hematopathology clinics worldwide.","abstract_has_math":false,"creators":["Cox, Andrew Michael"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Rajaram, Satwik","Huang, Lily","Danuser, Gaudenz","Mettlen, Marcel"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-06-03T19:48:35Z","date_published":"2025-06-03T19:48:35Z","updated_at":"2026-07-24T05:52:06Z","subjects":["Flow Cytometry","Machine Learning","Neoplasms","Neural Networks, Computer","Artificial Intelligence"],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["1522122341"],"render_values":[{"text":"1522122341","href":null,"code":true}]}]},"links":{"outbound_url":"https://hdl.handle.net/2152.5/10586","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Rajaram, Satwik","Huang, Lily","Danuser, Gaudenz","Mettlen, Marcel"]},{"key":"dc:creator","label":"Author","values":["Cox, Andrew Michael"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2025-06-03T19:48:35Z","2023-05","May 2023"]},{"key":"dc:type","label":"Dc Type","values":["Thesis","text"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Flow Cytometry","Machine Learning","Neoplasms","Neural Networks, Computer","Artificial Intelligence"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://hdl.handle.net/2152.5/10586","1522122341"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["This dissertation examines the use of modern machine learning (ML) methods to enhance the diagnostic workflow of hematological malignancies, particularly in cytogenetics and clinical flow cytometry. The recent advancements in the field of ML provide a unique opportunity to develop, adapt, and evaluate algorithms for augmenting the hematopathology diagnostic workflow. The field of hematology oncology involves a complex range of diseases, and the final diagnosis is reliant on the rapid analysis of several streams of biological data, including morphologic images, cytogenetic micrographs, molecular sequencing data, and high-dimensional multicolor flow cytometry (MFC). This dissertation explores the various strategies and advantages of using ML to augment this workflow, with a focus on automated screening for clinically informative recurring structural chromosomal abnormalities and serving as a clinical decision support system for clinical MFC data analysis. Most of the data used for this work are from patients with acute myeloid leukemia, but these ML algorithms can easily be extended to other indications. The ML models and techniques applied to cytogenetic micrographs and MFC data are predominantly of the deep learning variety, including convolutional neural networks (CNNs) and graph neural networks (GNNs), along with some other techniques like unsupervised ML, e.g., statistical representation learning and clustering, and graph partitioning. This research provides two key developments for the field. First, that CNNs can accurately detect over a dozen recurrent structural chromosome abnormalities common in several types of acute and chronic leukemia and simultaneously ignore non-recurrent structural abnormalities, laying the groundwork for an autonomous karyotyping system. Second, the first known application a GNN to MFC data is incredibly robust predicting various subtypes of acute promyelocytic leukemia and could be used as part of a clinical decision support system. Importantly, both these studies emphasize generalizability and explainabiliy. These approaches hold promise for facilitating automated analysis in a prospective randomized study and contribute to the digitization and ML revolution taking place in hematopathology clinics worldwide."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Machine Learning Techniques for Augmenting Hematological Malignancy Diagnosis"]}]}],"canonical_facts":{"dc:contributor":["Rajaram, Satwik","Huang, Lily","Danuser, Gaudenz","Mettlen, Marcel"],"dc:creator":["Cox, Andrew Michael"],"dc:date":["2025-06-03T19:48:35Z","2023-05","May 2023"],"dc:description":["This dissertation examines the use of modern machine learning (ML) methods to enhance the diagnostic workflow of hematological malignancies, particularly in cytogenetics and clinical flow cytometry. The recent advancements in the field of ML provide a unique opportunity to develop, adapt, and evaluate algorithms for augmenting the hematopathology diagnostic workflow. The field of hematology oncology involves a complex range of diseases, and the final diagnosis is reliant on the rapid analysis of several streams of biological data, including morphologic images, cytogenetic micrographs, molecular sequencing data, and high-dimensional multicolor flow cytometry (MFC). This dissertation explores the various strategies and advantages of using ML to augment this workflow, with a focus on automated screening for clinically informative recurring structural chromosomal abnormalities and serving as a clinical decision support system for clinical MFC data analysis. Most of the data used for this work are from patients with acute myeloid leukemia, but these ML algorithms can easily be extended to other indications. The ML models and techniques applied to cytogenetic micrographs and MFC data are predominantly of the deep learning variety, including convolutional neural networks (CNNs) and graph neural networks (GNNs), along with some other techniques like unsupervised ML, e.g., statistical representation learning and clustering, and graph partitioning. This research provides two key developments for the field. First, that CNNs can accurately detect over a dozen recurrent structural chromosome abnormalities common in several types of acute and chronic leukemia and simultaneously ignore non-recurrent structural abnormalities, laying the groundwork for an autonomous karyotyping system. Second, the first known application a GNN to MFC data is incredibly robust predicting various subtypes of acute promyelocytic leukemia and could be used as part of a clinical decision support system. Importantly, both these studies emphasize generalizability and explainabiliy. These approaches hold promise for facilitating automated analysis in a prospective randomized study and contribute to the digitization and ML revolution taking place in hematopathology clinics worldwide."],"dc:format":["application/pdf"],"dc:identifier":["https://hdl.handle.net/2152.5/10586","1522122341"],"dc:language":["en"],"dc:subject":["Flow Cytometry","Machine Learning","Neoplasms","Neural Networks, Computer","Artificial Intelligence"],"dc:title":["Machine Learning Techniques for Augmenting Hematological Malignancy Diagnosis"],"dc:type":["Thesis","text"]},"updated_at":"2026-07-24T05:52:06Z"}