Machine Learning applied to support medical decision in Transthoracic echocardiogram exams: A Systematic Review

The echocardiogram (ECHO) is an ultrasound of the heart used to diagnose heart diseases (DHC). The analysis and interpretation of ECHO are dependent on the doctor's experience. However, software that uses artificial intelligence to analyze ECHO images or videos is contributing to support the physician's decision. This paper aims to perform a Systematic Literature Review (SLR) on artificial intelligence (AI) techniques applied in the automation of Transthoracic Echocardiogram (TTE) processes, to support medical decisions. The study identified more than 800 articles on the topic in the leading scientific research platforms. To select the most relevant studies, inclusion and exclusion criteria were applied, where 45 articles were selected to compose the detailed study of the SRL. The results obtained with the extraction of information from the papers, identified 3 groups of primary studies, namely: identification of the cardiac vision plan, analysis of cardiac functions and detection of cardiac diseases. SRL identifies that the set of Machine learning (ML) techniques are being widely applied in the tasks of segmentation, detection and classification of images obtained from ECHO. The techniques based on Convolutional Neural Network (CNN), presented the best Accuracy rates. Research shows a strong interest in automating ECHO processes. However, it is still an open research field, with the potential to generate many publications for researchers.