Exploring Deep Learning and Machine Learning Approaches for Brain Hemorrhage Detection

Abstract

Brain hemorrhage is a critical medical condition that requires rapid and accurate diagnosis to minimize mortality 
and long-term neurological damage. Conventional diagnostic approaches rely on manual interpretation of CT scan 
images by radiologists, which can be time-consuming and may introduce variability due to human factors and limited 
availability of specialists. To address these challenges, this work presents a deep learning-based framework for 
automated brain hemorrhage detection from CT images. The proposed approach follows a structured pipeline that 
includes preprocessing, data augmentation, feature extraction, and classification. A custom Convolutional Neural 
Network (CNN) and a transfer learning-based InceptionV3 model are developed and evaluated on a dataset of 2,501 
CT images. The dataset consists of two classes (normal and hemorrhage) and is divided using a stratified approach 
to ensure balanced training. Experimental results indicate that while the custom CNN achieves strong performance, 
the InceptionV3 model demonstrates superior accuracy, robustness, and generalization capability. The best
performing model is further deployed as a Flask-based web application, enabling real-time predictions from uploaded 
CT images. These findings highlight the effectiveness of deep learning techniques as reliable decision-support tools 
for early diagnosis in clinical environments