On the Development of Artificial Intelligence-Assisted Auscultation
Dr Isabella Pak, Prof Joseph Suttie
Introduction
The use of Artificial Intelligence (AI) in the field of Cardiology is being increasingly investigated and developed for diagnostic and therapeutic purposes. Currently there is evidence that deep learning improves the accuracy and diagnostic capabilities of cardiac imaging including trans-thoracic echocardiogram (TTE), computed tomography coronary angiogram (CTCA), cardiac MRI and nuclear cardiac imaging. Recently, the application of such technology in digital stethoscopes is being used to improve auscultation, detect cardiac murmurs and predict cardiac failure. This review of existing literature aims to give an overview of current evidence for such use of AI.
Method/Description
A review of current studies into use of artificial intelligence in auscultation was conducted using PubMed and supplemented with other materials. 29 studies in total were found, and those that discussed only paediatric populations, or were study protocols were excluded.
Results
It was demonstrated across studies that cardiac murmurs were detectable using artificial intelligence and deep learning technology. Many of these studies found that digital stethoscopes and DL algorithms were able to accurately detect severe murmurs and cardiac disease, compared to clinicians. However, results in detection by algorithms were less accurate with milder cardiac disease. Additionally, highlighted in these studies was the need for normalisation across digital stethoscopes in order to further develop successful AI auscultation methods.
Conclusions
Deep learning has the potential to improve our ability to diagnose cardiac conditions through AI-assisted auscultation. This is an area that certainly requires further development in the hopes of improving safety and efficiency for patient care. This year, Wagga Wagga will be involved in an international trial of an FDA-approved digital stethoscope that employs artificial intelligence to detect low ejection fraction and predict cardiac failure.