Automatic calcium detection in echocardiography based on deep learning: A systematic review

Abstract

The diagnosis of many heart diseases involves the analysis of images from Computed Tomography (CT) or echocardiography, which is mainly done by a medical professional. By using Deep Learning (DL) algorithms, it is possible to create a data-driven tool capable of processing and classifying this type of image, to support physicians in their tasks, improving healthcare efficiency by offering faster and more accurate diagnoses. The aim of this paper is to perform a systematic review on DL uses for automated methods for calcium detection, identifying the state of this art. The systematic review was based on PRISMA methodology to identify relevant articles about image processing using Convolutional Neural Networks (CNN) in the cardiac health context. This search was conducted in Scopus and Web of Science Core Collection, and the keywords considered included (1) Deep Learning, (2) Calcium Score, (3) CT-Scan, (4) Echocardiography. The review yielded 82 research articles, 38 of which were in accordance with the initial requirements by referring to image processing and calcium score quantification using DL models. DL is reliable in the implementation of classification methods for automatic calcium scoring. There are several developments using CT-Scan, and a need to replicate such methods to echocardiography.