Determination of vascular stenosis on angiography images using convolutional neural network method

Coronary artery disease (CAD) is the most common type of heart disease worldwide. Cardiovascular diseases usually refer to conditions that include narrowed or blocked blood vessels that can cause heart attacks, angina or stroke. Invasive coronary angiography (ICA) is the standard clinical method for identifying coronary arteries and is currently the gold standard for CAD diagnosis. ICA is the X-ray imaging of cardiac cavities and coronary arteries using contrast agent. Computer aided detection systems are very important in terms of supporting physicians' decision making. In this thesis, a method was developed to analyze angiography images using convolutional neural network (CNN). In order to improve the accuracy of the method, cardiovascular vessels were first segmented by classical methods presented in the literature and these images were evaluated with the CNN algorithm. The developed method was tested on the cases obtained from the databases containing the images scored by the physicians as open source and 94.84% accuracy was achieved.