CARDIOVASCULAR RISK IN PATIENTS WITH TAKAYASU ARTERISTIS USING DEEP GATED ATTENTION BAG LEVEL CLASSIFIER

Abstract

Takayasu Arteritis (TA) being a chronic inflammatory state chiefly influences the aorta and its crucial branches, resulting in severe Cardio Vascular Disease (CVD) impediments. TA-CVD detection employing image analysis is more complex and hence to design a Deep Learning based computer aided solution is indispensable to address this demanding issue. To focus on this issue, a Gated Attention Bag Level Classifier (GA-BLC) design that can automatically extract representative features from TA images for identifying cardiovascular risk in patients. GA-BLC method consists of three stages, pre-processing using Deep Learning Reconstruction Weighted Contrast-enhanced Pre-processing model, CutMixaugmentationandResNet Gated Attention Bag-level Classifier for Cardiovascular Prediction in Patients with Takayasu Arteristis. To evaluate the performance of obtained deep features, bag-level classifier is applied. The experimental results for clinical data with TA along with training images show that the deep features with bag-level classifier achieve the highest performances 13% accuracy and 25% accuracy compared to the conventional classifier. Moreover the inter-reliability and intra-reliability were found to be improved by 14% and 18% compared to the traditional classifier. The findings put forward that our proposed method has competence to locate biomarkers for diagnosis of cardiovascular risk in patients with TA, which will aid in the evolution of computer assisted diagnostic method by specialists.