Cardiac Diagnosis Classification Using Deep Learning Pipeline on Apache Spark

Abstract

Medical data are heterogeneous and complex data that are difficult to analyze and manage with traditional software or hardware. Deep learning is generating a major impact on medical imaging and ECG image analysis is more popular than the analysis of ECG’ signals processing. So, this paper proposes the cardiac diagnosis classification system using the transfer learning of deep learning pipeline on Apache Spark for the classification of ECG images. The deep learning pipeline enables fast transfer learning as a research problem of machine learning that focuses on storing knowledge gained on unsupervised segmented ECG images. To get the correct classification of heart diagnoses, the system needs to segment the ECG images and uses the principal component analysis to reduce unsupervised segmented images and select the sample diagnosis images from ECG segmented images. These segmented images are the high dimension of heterogeneous phenotypes. The proposed system classifies the five sample images of cardiac diagnosis by combining with the DL pipeline’s Convolutional Neural Network (InceptionV3) and Logistic Regression. So, the system uploads the scanning images that are segmented into High Distributed File System (HDFS) using the apache spark framework. This paper proposes a cardio diagnosis detection system for efficient classification using CNN on extracting unsupervised data features of health care.