Intracranial Brain Haemorrhage Segmentation and Classification
Keywords:Classification, Detection, Faster R-CNN, FCN, ICH, Segmentation, U-NET
Traumatic brain injuries are categorized as sudden damage to the brain which may be caused by a blow to the head. A traumatic brain injury can cause intracranial bleeding which may lead to Intracranial hemorrhage (ICH). Computerized Tomography (CT) scans are widely used by radiologists in the detection and diagnosis of ICH. A CT scan creates images of the brain which can help detect bleeding and other signs of trauma to the head. However, accurate detection and diagnosis of ICH depend on access to an experienced radiologist. Failure to accurately detect and treat ICH promptly can lead to disability or even death. This project aims to develop an artificially intelligent system capable of detecting, diagnosing ICH, and classifying its sub-types. For this purpose, we will employ the techniques of computer vision and machine learning to train a Fully Convolutional Network (FCN) called u-net on a publicly available data set of head CT scans. The development process will include taking CT scans as input, using u-net as an FCN to perform semantic segmentation to classify the type of ICH, and the region of the brain affected by it. The proposed system will facilitate junior doctors and radiologists by providing them with assistance in the detection of ICH and its subtypes.