Detection and classification of cerebral infarction using artificial intelligence techniques

Cerebral infarction, also known as ischemic stroke, occurs when blood flow to brain tissue is reduced or completely cut off due to a blockage in the brain vessels. Impaired blood flow deprives brain cells of oxygen and vital nutrients, which can lead to permanent brain damage. In our country, the process of patients recognizing the symptoms of infarction, presenting to the hospital, initial evaluation in the emergency department, suspicion of infarction and subsequent diagnosis with Magnetic Resonance Imaging (MRI) can be prolonged for various reasons. This poses a serious risk for vital clinical conditions such as cerebral infarction where rapid and accurate diagnosis is crucial. In this context, computer-aided diagnosis systems aim to assist radiologists in diagnostic processes by contributing to faster and more accurate interpretation of medical images. In this thesis, automatic detection of cerebral infarction on MRI images will be realized with artificial intelligence based methods. In the first stage, Convolutional Neural Networks (CNN) architecture will be used to classify the infarct regions. Then, U-Net based semantic segmentation will be applied to precisely segment the infarct areas on the image. The decision support system to be developed with this approach aims to both accelerate clinical processes and contribute to the literature in the field of medical imaging.