This document presents a method for automatically classifying CT brain images according to different types of head trauma. The method involves three main steps: 1) preprocessing images to segment potential hemorrhage regions, 2) extracting features from each region like size, shape, location, 3) classifying each region and overall image using machine learning. The method was tested on 35 CT brain images and achieved an average accuracy of 93% in classifying potential hemorrhage regions into categories like epidural hemorrhage, subdural hemorrhage, and intracerebral hemorrhage.