JMIR Publications

ABSTRACT

Background:

Cardiovascular disease remains the leading cause of mortality worldwide, with cardiac fibrosis contributing significantly to its pathophysiology. Identifying cardiac fibrosis is essential for prognosis and management, yet current diagnostic methods face challenges due to invasiveness, cost, and inaccessibility. There is a pressing need for novel and accessible techniques for detecting cardiac fibrosis. The electrocardiogram (ECG) presents a potential solution, using machine learning (ML) to infer fibrosis from changes in electrophysiology.

Objective:

This review explores the historical use of ECG technology for detecting cardiac fibrosis and the role of ML in enhancing ECG's diagnostic capabilities for cardiac fibrosis.

Methods:

Fibrosis ECG electrophysiology and detection using non-ML were explored using a narrative review of the literature. Additionally, we systematically reviewed studies that applied ML to ECG data to detect cardiac fibrosis. All searches were conducted in PubMed, with eligibility assessment and data extraction completed by two reviewers.

Results:

Visual ECG evaluation techniques for fibrosis include identifying fragmented QRS complexes and Selvester scoring, though these are burdensome and have limited accuracy. Several studies applied ML to ECG data for fibrosis detection. Notable advancements were found, with some studies achieving high sensitivity and specificity. However, existing research limitations include small sample sizes, lack of randomized controlled trials, and limited external validation. Variability in data preprocessing and feature engineering methods may also impact reproducibility and generalizability.

Conclusions:

Despite the potential of ECG-ML approaches for detecting cardiac fibrosis, further research is needed to address current limitations and improve model robustness. Larger, diverse studies with standardized methodologies are essential to fully harness the benefits of ECG-ML for cardiac fibrosis detection.