A Comprehensive Review on Deep Learning Approach for Intracranial Haemorrhage Detection and Analysis of Multimodal Dataset

Abstract

The process of creating visual representations of every internal part and constructing the physiological functions of the body as well as organs functioning is known as medical imaging. For medical support and therapy, the medical pictures are retrieved using a variety of procedures, including MRI (Magnetic Resonant Imaging) and CT (Computed Tomography). Intracranial hemorrhages (ICH) can result from severe brain injury. The term "intracranial hemorrhage" refers to a blood clot inside the skull. If it has not been adequately detected, this disease may cause considerable impairment or maybe death. The proposed study aims to analyze the supervised DL models to determine whether there is a hemorrhage in CT brain images.Convolution neural networks are particularly helpful for identifying patterns in images and for enabling scenarios to contain objects and faces. In the first phase, a pre-processing step is performed. The images endure pre-processing procedures to prepare them for additional processing. CNN eliminates the requirement for manual extraction of features by classifying pictures directly from image data using patterns or samples. This method has the benefit of producing findings with a clear range of errors and is extremely efficient in high-dimensional settings. This article demonstrates the processing of CT brain images for cerebral bleeding identification using a variety of approaches including region-based growth. This research highlights the superior hemorrhage results in DL models, including GoogleNet, ResNet-50, and AlexNet methods. These methods offer more precise findings for detecting ICH on CT scans and when analyzing multimodal medical imaging datasets.