CT Analysis of Acute Ischemic Stroke

In this project data is collected containing CT scans and CT angiography images for segmentation, registration and masking purposes that can be used to diagnose ischemic stroke. A U-Net based deep learning model is used to obtain segmentation maps of the brain from its axial CT scans, which are used with a window frame to select the relevant area. Then a multiplier between the image and the mask is used to obtain the brain segment itself. After co-registering the images, these segments are further segmented to obtain vessel segments using thresholding, which can be used for collateral scoring and 3D construction. Another approach is that of Digital Subtraction Angiography: noncontrast and normal CTs are registered and digitally subtracted to give a direct image of the vessels. This project however, focuses solely on establishing a foundation for stroke diagnoses by focusing on vessel segmentation from segmented brain slices, therefore the study is focused on healthy patients.