News: AI has potential to improve diagnostic accuracy, earlier detection of cardiac conditions, analysis finds

A review of 800 studies found that electrocardiogram (ECG) interpretation assisted by artificial intelligence (AI) has the potential to improve diagnostic accuracy and enable earlier detection of cardiac conditions, particularly in resource-limited outpatient settings, JustCoding reported.

The review, published by Cureus, aimed to explore the effectiveness of using AI to improve interpretation of the 12-lead ECG, which is a diagnostic tool for early detection of cardiac abnormalities. The authors of the review note that these ECGs are widely available and inexpensive compared to other diagnostic modalities and can prevent adverse outcomes when interpreted correctly and promptly, but previous reports have noted “low proficiency among trainees, as well as inconsistencies in algorithm-based diagnoses due to non-standardized criteria.” And these deficiencies manifest more in outpatient and rural settings due to limited access to specialists, which puts more reliance on “potentially flawed automated interpretations.”

For the review, authors reviewed published studies on AI-assisted ECG interpretations in outpatient settings and how they affected diagnostic accuracy and clinical outcomes. Overall, the studies demonstrated improvements in diagnostic accuracy and clinical outcomes, as well as reductions in human error, through either deep learning models or AI algorithms.

Several studies compared AI performance with standard diagnostic techniques and found that AI models performed just as well or outperformed traditional diagnostic methods and expert clinicians. This was particularly true for AI models using deep learning. AI was also found to be able to detect cardiac conditions long before conventional diagnoses are possible. For example, one study examined a convolutional neural network that outperformed expert cardiologists by predicting disease onset nearly a year in advance.

The authors noted that despite AI’s capacity to enhance diagnostic accuracy and enable earlier detection of cardiac conditions, there are still recurring issues with its use. For example, some AI models tend to over diagnose.

Editor’s note: This article originally appeared in JustCoding. To read the full analysis from Cureus, click here.