JMIR Publications

ABSTRACT

Background:

At present, the development of artificial intelligence is rapid. We have noticed that the artificial intelligence based on MRI is controversy in diagnosing myocarditis.

Objective:

The aim is to assess the diagnostic capability of artificial intelligence (AI) in identifying myocarditis through cardiovascular magnetic resonance imaging (MRI)

Methods:

A comprehensive search of studies was conducted through Web of Science, Embase and PubMed with a focus on researches published before June 7, 2026. If the studies assessment involved the application of AI models based on cardiovascular MRI in the detection of myocarditis, it will be included. The bivariate random effects model was used to ascertain the joint consideration of sensitivity and specificity. Heterogeneity across studies was assessed using the I² statistic. Employing the revised QUADAS-2 tool assesses the risk of bias. The certainty of evidence was evaluated according to GRADE framework.

Results:

Out of the initially identified 1,222 studies, there eventually included 17 studies. The ultimate analysis involved 93,740 patients and images. For myocarditis, AI showed that the sensitivity was 0.93 (0.88 − 0.96) and specificity was 0.94 (0.89 − 0.97), with the AUC of 0.98 (0.96 - 0.99). The asymmetry test of the Deeks' funnel plot did not indicate any significant publication bias (P = 0.46). Meta-regression and subgroup analysis revealed that there are markedly different in groups of analysis, AI method, reference standard and years (P < 0.05).

Conclusions:

By aggregating the data, this meta-analysis manifested that cardiovascular MRI based on AI revealed excellent ability in diagnosing myocarditis. However, this study is subject to limitations, including its retrospective design and the methodological heterogeneity across cardiovascular MRI. In the future, there is an urgent need for more forward-looking multi-center studies to prove this conclusion.