Calcified Tissue Detection using Deep Learning Models in Intravascular Ultrasound images

Cardiovascular disease is the world's biggest cause of mortality. Heart disease is thought to have calcification as a significant contributing component. Currently, intravascular ultrasonography (IVUS) pictures are used to visually check for the presence of calcification by medical professionals. The goal of the study is to identify dense calcification using IVUS images taken at a 40MHz frequency. To detect these tissue frames, the features were extracted using deep learning models namely, AlexNet, GoogLeNet, SqueezeNet architectures. The IVUS pullbacks of four real patients were used in the tests. According to experimental findings, AlexNet, a deep learning architecture that has been trained, performs better than the other two models. The classification accuracy of the AlexNet (98.10%) model significantly outperforms that of two other previously trained models, GoogLeNet (89.93%) and SqueezeNet (92.80%). Future research into other feature selection techniques like ReliefF, PSO, ACO, and others, together with other models like Vision Transformers, could enhance the diagnostic process' overall accuracy.