Intracranial hemorrhage detection and classification from CT images based on multiple features and machine learning approaches

Abstract

The regulating organ of the body is the brain. Early diagnosis of brain disorders can have a significant impact on efforts to treat them. A brain hemorrhage is a form of stroke caused by a bursting artery in the brain, resulting in bleeding in the surrounding tissues. Through a brain Computed Tomography (CT) scan, brain hemorrhage can be identified. CT is the most extensively used diagnostic imaging technology for identifying brain illnesses due to its speed, low cost, and wide variety of uses. During a CT scan, a small X-ray beam revolves around the body to capture a sequence of images from different angles. The computer then produces a cross-sectional representation of the body. Intracranial hemorrhage (ICH) is a medical condition that requires prompt identification and treatment. Since ICH early detection and therapy can improve health outcomes, there is a need for a triage system that can immediately identify and speed up the treatment process. In this paper, we will use standard machine learning (Support Vector Machine, Random Forest and Decision Tree) methodologies to present a method for automatically detecting the ICH in a two-dimensional reduced form of a CT scan of the brain. Four main steps make up the method. First, a preprocessing pipeline that can successfully remove the bone from the skull is put into place. The following step is applying a feature extraction method. Then, a suitable feature-selection (PCA) model is proposed, which will enhance the model's performance by minimizing any redundancy produced by the selected feature extraction. The data set from the CT scans is classified into normal and abnormal in the last stage, which involves training and testing a machine learning model. The accuracy for our proposed model using Random Forest (RF), is 92.5%. RF achieves higher performance than other used ML methods.