Detecting Left Heart Failure in Echocardiography through Machine Learning: A Systematic Review

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Heart failure presents a significant healthcare challenge, particularly in Asia.
• Echocardiography is vital for assessing cardiac function, but faces resource and time constraints, exacerbated by the COVID-19 pandemic.
• Artificial intelligence (AI) offers potential for accurate and rapid heart failure diagnosis.

Purpose of the Study:

• To systematically review the literature on the application of AI in echocardiography for heart failure diagnosis.
• To assess the accuracy and methodologies of AI-driven diagnostic tools.

Main Methods:

• Systematic literature search across multiple databases (Europe PMC, ProQuest, Science Direct, PubMed, IEEE) adhering to PRISMA guidelines.
• Quality and risk of bias assessment of included studies using QUADAS-2.
• Analysis of 14 selected studies out of 2105 retrieved.

Main Results:

• Fourteen studies were included, with five showing risks of bias.
• Commonly used echocardiography views were apical four-chamber (A4C) and apical two-chamber (A2C) from 2D and 3D datasets.
• Convolutional neural networks were the most frequent AI method, with diagnostic accuracy ranging from 57% to 99.3%.

Conclusions:

• AI applications in echocardiography demonstrate potential for enhanced and expedited diagnosis of left heart failure.
• AI should be viewed as a complementary tool to support clinicians, not replace them.
• Clinician involvement remains indispensable for accurate diagnosis and comprehensive patient management.