Applications of machine learning in glioblastoma diagnosis, classification, treatment, and prognosis
Advisor(s)
Dr. Jane Wu, Northwestern University Feinberg School of Medicine
Warren McGee, Northwestern University Feinberg School of Medicine
Location
Room A119
Start Date
26-4-2019 11:25 AM
End Date
26-4-2019 11:40 AM
Abstract
Glioblastomas are highly invasive, malignant, grade IV astrocytomas, formed primarily from cancerous astrocytes and sustained by intense angiogenesis, often causing non-specific symptoms and creating difficulty for definitive diagnoses. This study aims to utilize artificial intelligence, machine learning, and deep learning techniques in order to provide an accurate molecular classification and survival prognosis for glioblastoma patients using magnetic resonance imaging, clinical, and genomic data. Images from TCIA-TCGA and IvyGAP datasets will be processed and used to train and test computer algorithms. At the study’s current stage, raw data has been processed for use and an algorithm is in development for the aforementioned purposes. Criteria are also being determined for selecting data of utility to the current study. A multilayer perceptron and a convoluted network will be combined in a single end-to-end Keras model designed to accept mixed data inputs from processed clinical, genomic, and MR imaging data. These results can help identify predictive features that could assist in providing more accurate and comprehensive diagnoses and that are significant for survivability.
Applications of machine learning in glioblastoma diagnosis, classification, treatment, and prognosis
Room A119
Glioblastomas are highly invasive, malignant, grade IV astrocytomas, formed primarily from cancerous astrocytes and sustained by intense angiogenesis, often causing non-specific symptoms and creating difficulty for definitive diagnoses. This study aims to utilize artificial intelligence, machine learning, and deep learning techniques in order to provide an accurate molecular classification and survival prognosis for glioblastoma patients using magnetic resonance imaging, clinical, and genomic data. Images from TCIA-TCGA and IvyGAP datasets will be processed and used to train and test computer algorithms. At the study’s current stage, raw data has been processed for use and an algorithm is in development for the aforementioned purposes. Criteria are also being determined for selecting data of utility to the current study. A multilayer perceptron and a convoluted network will be combined in a single end-to-end Keras model designed to accept mixed data inputs from processed clinical, genomic, and MR imaging data. These results can help identify predictive features that could assist in providing more accurate and comprehensive diagnoses and that are significant for survivability.