Evaluating the Impact of Segmentation on Feature Extraction for Multi-Modal Brain Tumor Classification
Iris Ella Cruz, Paige Camaya, Dona Inayyah Don Nazwim
Department of Biology
Faculty Supervisor: Sara El Alaoui
Our goal is to develop a multi-modal, multi-class classification model that not only detects the presence of a tumor but also classifies the tumor sub-regions—Necrotic Core (1), Edema (2), and Enhancing Tumor (4)—against non-tumor regions (0). We compare two approaches: segmentation-based feature extraction, where features are derived from identified regions of interest (ROIs), and whole-image feature extraction, where features are extracted without segmentation.