Brain Tumor Radiogenomic Classification Using Deep Learning Algorithms

Brain tumor pathology must be determined using medical imaging to make a timely diagnosis. Brain tumors can be screened using a variety of medical imaging techniques. The current method of genetic analysis takes a long time and requires the surgical removal of samples of brain tissue. Analyzing tumor DNA from a biopsy or surgical resection reveals the degree of O-6-methylguanine-DNA methyltransferase (MGMT) promoter methylation. MGMT promoter methylation is currently used as a prognostic and predictive marker, although intra-tumoral heterogeneity of methylation can obstruct whole-tumor characterization, leading to a range of survival outcomes. It has been established that the effectiveness of alkylating medicines, including Temozolomide (TMZ), the most popular chemotherapy for glioblastoma, depends on the degree of MGMT promoter methylation. Because of this, MGMT promoter expression epigenetic suppression has been used as a crucial molecular marker in therapeutic practice. Glioblastoma patients with a high level of MGMT promoter methylation are more responsive to TMZ and have a better overall survival (OS) compared to those with a low level of MGMT promoter methylation. In this study, ResNet-50 and convolutional neural network (CNN) deep learning algorithms were used to predict the importance of MGMT promoter methylation. This study involved data acquisition, exploratory data analysis (EDA), image preprocessing, and performance evaluation of the suggested models. The accuracy was determined to assess how well the MGMT prediction performed. ResNet-50 and CNN, two commonly used deep learning algorithms, have accuracy rates of 86% and 85%, respectively. As a result, ResNet-50 exhibits a high degree of precision in MGMT promoter methylation prediction.