Title : Gliosmart: A novel, non-invasive-personalized treatment response for Glioblastoma Multiforme (GBM)
Abstract:
Glioblastoma multiforme (GBM), a deadly brain cancer with a median survival of only 15-18 months, presents a significant hurdle for personalized treatment. Accounting for nearly half of primary central nervous system tumors, GBM is conventionally treated through surgical resection, radiation therapy, and temozolomide (TMZ) chemotherapy. The MGMT gene's promoter methylation, observed in 40-60% of glioblastomas, enhances the TMZ response and serves as a prognostic biomarker. That is why the presence of MGMT promoter can be used to determine which treatment method to use for certain tumors, based on whether TMZ will be effective or not. However, the current gold-standard of genetic analysis to determine MGMT methylation status from surgical specimens is time consuming and may necessitate subsequent surgeries based on the results. This research introduces an innovative machine learning solution to streamline MGMT-status determination. Implemented using a convolutional neural network architecture for tumor-identification and predicting MGMT methylation, the model utilizes MRI brain scan images of GBM patients from The Cancer Imaging Archive-(TCIA) and genomic data from The Cancer Genome Atlas-(TCGA), achieving an impressive 96%accuracy. To further demonstrate clinical viability, I integrated the model with a web-application, "GlioSmart," enabling rapid MRI scan uploads and delivering MGMT status in less than 2 seconds. This novel breakthrough minimizes the need for invasive biopsies, significantly reducing time and cost. Oncologists can leverage GlioSmart to recommend personalized treatment plans based on MGMT status. Beyond overcoming biopsy limitations, this innovative AI driven approach revolutionizes treatment decision making, offering more effective and personalized care for GBM patients.
