Self-attention based BiLSTM-CNN classifier for the prediction of ischemic and non-ischemic cardiomyopathy

Abstract

Approximately 26 million individuals are suffering from heart failure, according to the global annual report. Despite higher inter-rater variability, endomyocardial biopsy (EMB) is still regarded the gold standard for assessing heart failure. Therefore, we proposed and implemented a new unified architecture consist of convolutional layers, bidirectional LSTM (BiLSTM), and self-attention mechanism to predict the ischemic and non-ischemic cardiomyopathy using histopathological images. The proposed model is based on self-attention that implicitly focus to the information outputted from the hidden layers of BiLSTM. Through our results we demonstrate that this framework carries high learning capacity and is able to improve the classification performance.