Integrating Active Contours and Decision Trees for Systematic Hemorrhage Detection

Medical imaging is a pivotal tool for visualizing internal bodily structures and comprehending organ and tissue functions. This study focuses on a system designed to identify and categorize hemorrhages in CT scans. Different imaging techniques like CT scans, MRI, and X-rays can produce these visuals. The system's main objective is to identify hemorrhages and, when found, categorize them. To accomplish this, various initial processing steps, such as converting to grayscale, adjusting size, detecting edges, and enhancing clarity, are executed on the CT images, readying them for further evaluation. Morphological operations identify shape-related hemorrhage features, using the Sobel and marker techniques to highlight regions of interest. Subsequently, the watershed algorithm facilitates hemorrhage detection through segmentation. Once hemorrhages are detected, feature extraction is performed for classification. Active contours define hemorrhage areas, and extracted features enter a Decision Tree classifier for type determination. Results can be interpreted in collaboration with medical professionals. This research enhances hemorrhage prediction and classification in medical images, achieving 92% average accuracy in classifying three hemorrhage types.