Application of deep learning method to spect images forevaluation of ischemia in coronary artery disease

Coronary artery disease (CAD) is one of the causes of abnormal functioning of the heart, which is one of the vital organs. The occurrence of ischemia in the cardiovascular system as a result of narrowing-occlusion of the coronary artery vessels that provide adequate nutrition and oxygenation of the heart is called CAD. SPECT (Single Photon Emission Computed Tomography) imaging technique, which allows functional evaluation of the cardiovascular system, is one of the frequently preferred nuclear medicine applications in the diagnosis of CAD. By comparing stress-resting SPECT images, it provides information about the function and risk status of the heart, and the location of the narrowing-occlusion in the coronary arteries. In the current situation, the information obtained from the SPECT images is interpreted by the experts and the disease is diagnosed. In this study, a system that is capable of meeting the processing of knowledge that is increasing day by day has been proposed. Convolutional neural network (ESA) algorithms and deep learning (DL) model designed by including powerful hardware technologies can help experts in diagnosing CAD. The training and testing of the SPECT MPI image dataset was carried out on various transfer learning (BL) architectures, and ischemia/living tissue evaluation was performed. As a result of the study, when the performance performance criteria of the pre-trained DenseNet121, InceptionResNetV2, InceptionV3, MobileNet, MobileNetV2, ResNet50, VGG16, VGG19, Xception architectures are examined, the highest training accuracy is InceptionV3, with the highest accuracy %98, sensitivity %98, F1-score %98 and AUC value of 0.977 were analyzed in the Xception model. In the test data set, the lowest accuracy value was calculated for VGG16 %98.3, VGG19 %85 and MobileNetV2 %77.2, while the remaining models achieved %100 accuracy. Compared to other studies in the literature, it is aimed to evaluate ischemia/living tissue with the DS method, which has the ability to analyze based on SPECT MPI images.