The overall aggressiveness of a glioma is measured by histologic and molecular analysis of tissue samples. However, the well-known spatial heterogeneity in gliomas limits the ability for clinicians to use that information to make spatially specific treatment decisions. Magnetic resonance imaging (MRI) visualizes and assesses the tumor. But, the exact degree to which MRI correlates with the actual underlying tissue characteristics is not known.
In this work, we derive quantitative relationships between imaging and underlying pathology. These relations increase the value of MRI by allowing it to be a better surrogate for underlying pathology and they allow evaluation of the underlying biological heterogeneity via imaging. This provides an approach to answer questions about how tissue heterogeneity can affect prognosis.
We estimated the local pathology within tumors using imaging data and stereotactically precise biopsy samples from an ongoing clinical imaging trial. From this data, we trained a random forest model to reliably predict tumor grade, proliferation, cellularity, and vascularity, representing tumor aggressiveness. We then made voxel-wise predictions to map the tumor heterogeneity and identify high-grade malignancy disease.
Next, we used the previously trained models on a cohort of 1,850 glioma patients who previously underwent surgical resection. High contrast enhancement, proliferation, vascularity, and cellularity were associated with worse prognosis even after controlling for clinical factors. Patients that had substantial reduction in cellularity between preoperative and postoperative imaging (i.e. due to resection) also showed improved survival.
We developed a clinically implementable model for predicting pathology and prognosis after surgery based on imaging. Results from imaging pathology correlations enhance our understanding of disease extent within glioma patients and the relationship between residual estimated pathology and outcome helps refine our knowledge of the interaction of tumor heterogeneity and prognosis.
","abstract_html":"<p>The overall aggressiveness of a glioma is measured by histologic and molecular analysis of tissue samples. However, the well-known spatial heterogeneity in gliomas limits the ability for clinicians to use that information to make spatially specific treatment decisions. Magnetic resonance imaging (MRI) visualizes and assesses the tumor. But, the exact degree to which MRI correlates with the actual underlying tissue characteristics is not known.</p> <p>In this work, we derive quantitative relationships between imaging and underlying pathology. These relations increase the value of MRI by allowing it to be a better surrogate for underlying pathology and they allow evaluation of the underlying biological heterogeneity via imaging. This provides an approach to answer questions about how tissue heterogeneity can affect prognosis.</p> <p>We estimated the local pathology within tumors using imaging data and stereotactically precise biopsy samples from an ongoing clinical imaging trial. From this data, we trained a random forest model to reliably predict tumor grade, proliferation, cellularity, and vascularity, representing tumor aggressiveness. We then made voxel-wise predictions to map the tumor heterogeneity and identify high-grade malignancy disease.</p> <p>Next, we used the previously trained models on a cohort of 1,850 glioma patients who previously underwent surgical resection. High contrast enhancement, proliferation, vascularity, and cellularity were associated with worse prognosis even after controlling for clinical factors. Patients that had substantial reduction in cellularity between preoperative and postoperative imaging (i.e. due to resection) also showed improved survival.</p> <p>We developed a clinically implementable model for predicting pathology and prognosis after surgery based on imaging. Results from imaging pathology correlations enhance our understanding of disease extent within glioma patients and the relationship between residual estimated pathology and outcome helps refine our knowledge of the interaction of tumor heterogeneity and prognosis.</p>","abstract_has_math":false,"creators":["Gates, Evan","0000-0002-0829-6286
"],"institution":null,"degree_name":"Doctor of Philosophy (PhD)","degree_level":"Dissertation (PhD)","degree_discipline":null,"degree_department":null,"school":null,"contributors":["David Fuentes, PhD","Dawid Schellingerhout, MD","Kristy Brock, PhD"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2021,"date_issued":"2021-05-01T07:00:00Z","date_published":"2021-05-01T07:00:00Z","updated_at":"2026-07-24T05:49:16Z","subjects":["glioma","pathology","biopsy","machine learning","magnetic resonance imaging","MRI","neurosurgery","Diagnosis","Medical Pathology","Medicine and Health Sciences","Neoplasms","Radiology"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://digitalcommons.library.tmc.edu/utgsbs_dissertations/1083","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["David Fuentes, PhD","Dawid Schellingerhout, MD","Kristy Brock, PhD"]},{"key":"dc:creator","label":"Author","values":["Gates, Evan","0000-0002-0829-6286
"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2021-05-03T07:00:00Z"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Dissertation (PhD)"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["glioma","pathology","biopsy","machine learning","magnetic resonance imaging","MRI","neurosurgery","Diagnosis","Medical Pathology","Medicine and Health Sciences","Neoplasms","Radiology"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://digitalcommons.library.tmc.edu/utgsbs_dissertations/1083"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["The overall aggressiveness of a glioma is measured by histologic and molecular analysis of tissue samples. However, the well-known spatial heterogeneity in gliomas limits the ability for clinicians to use that information to make spatially specific treatment decisions. Magnetic resonance imaging (MRI) visualizes and assesses the tumor. But, the exact degree to which MRI correlates with the actual underlying tissue characteristics is not known.
In this work, we derive quantitative relationships between imaging and underlying pathology. These relations increase the value of MRI by allowing it to be a better surrogate for underlying pathology and they allow evaluation of the underlying biological heterogeneity via imaging. This provides an approach to answer questions about how tissue heterogeneity can affect prognosis.
We estimated the local pathology within tumors using imaging data and stereotactically precise biopsy samples from an ongoing clinical imaging trial. From this data, we trained a random forest model to reliably predict tumor grade, proliferation, cellularity, and vascularity, representing tumor aggressiveness. We then made voxel-wise predictions to map the tumor heterogeneity and identify high-grade malignancy disease.
Next, we used the previously trained models on a cohort of 1,850 glioma patients who previously underwent surgical resection. High contrast enhancement, proliferation, vascularity, and cellularity were associated with worse prognosis even after controlling for clinical factors. Patients that had substantial reduction in cellularity between preoperative and postoperative imaging (i.e. due to resection) also showed improved survival.
We developed a clinically implementable model for predicting pathology and prognosis after surgery based on imaging. Results from imaging pathology correlations enhance our understanding of disease extent within glioma patients and the relationship between residual estimated pathology and outcome helps refine our knowledge of the interaction of tumor heterogeneity and prognosis.
"]},{"key":"dc:title","label":"Title","values":["Imaging Based Prediction of Pathology In Adult Diffuse Glioma With Applications to Therapy and Prognosis"]}]}],"canonical_facts":{"dc:contributor":["David Fuentes, PhD","Dawid Schellingerhout, MD","Kristy Brock, PhD"],"dc:creator":["Gates, Evan","0000-0002-0829-6286
"],"dc:date.available":["2021-05-03T07:00:00Z"],"dc:description.abstract":["The overall aggressiveness of a glioma is measured by histologic and molecular analysis of tissue samples. However, the well-known spatial heterogeneity in gliomas limits the ability for clinicians to use that information to make spatially specific treatment decisions. Magnetic resonance imaging (MRI) visualizes and assesses the tumor. But, the exact degree to which MRI correlates with the actual underlying tissue characteristics is not known.
In this work, we derive quantitative relationships between imaging and underlying pathology. These relations increase the value of MRI by allowing it to be a better surrogate for underlying pathology and they allow evaluation of the underlying biological heterogeneity via imaging. This provides an approach to answer questions about how tissue heterogeneity can affect prognosis.
We estimated the local pathology within tumors using imaging data and stereotactically precise biopsy samples from an ongoing clinical imaging trial. From this data, we trained a random forest model to reliably predict tumor grade, proliferation, cellularity, and vascularity, representing tumor aggressiveness. We then made voxel-wise predictions to map the tumor heterogeneity and identify high-grade malignancy disease.
Next, we used the previously trained models on a cohort of 1,850 glioma patients who previously underwent surgical resection. High contrast enhancement, proliferation, vascularity, and cellularity were associated with worse prognosis even after controlling for clinical factors. Patients that had substantial reduction in cellularity between preoperative and postoperative imaging (i.e. due to resection) also showed improved survival.
We developed a clinically implementable model for predicting pathology and prognosis after surgery based on imaging. Results from imaging pathology correlations enhance our understanding of disease extent within glioma patients and the relationship between residual estimated pathology and outcome helps refine our knowledge of the interaction of tumor heterogeneity and prognosis.
"],"dc:identifier":["https://digitalcommons.library.tmc.edu/utgsbs_dissertations/1083"],"dc:subject":["glioma","pathology","biopsy","machine learning","magnetic resonance imaging","MRI","neurosurgery","Diagnosis","Medical Pathology","Medicine and Health Sciences","Neoplasms","Radiology"],"dc:title":["Imaging Based Prediction of Pathology In Adult Diffuse Glioma With Applications to Therapy and Prognosis"],"thesis:degree_level":["Dissertation (PhD)"],"thesis:degree_name":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-24T05:49:16Z"}