Deep Learning for Cerebral Hemorrhage Detection and Classification in Head CT Scans Using CNN

Abstract

The classification of cerebral hemorrhages in Computed Tomography (CT) scans using Convolutional Neural Networks (CNNs) is a rapidly evolving field in medical imaging. Brain hemorrhages are a critical condition that can result in serious health consequences and death. Recently, deep neural networks have been employed for image identification and classification, producing encouraging outcomes in medical image analysis. The objective of this study is to utilize deep learning methods and CNNs to identify brain hemorrhages in CT images. The inspiration for this research stems from the challenges faced by physicians in accurately recognizing brain hemorrhages, especially in the early stages when misdiagnosis is more likely. Through a series of CT experiments, two pretrained CNNs (VGG16 and VGG19) were developed and evaluated for image categorization as either hemorrhage or non-hemorrhage. The VGG16 pre-trained model showed exceptional accuracy compared to the VGG19 model. The VGG16 model also achieved the highest accuracy of 99.10% compared to all reference studies.