A systematic integrity audit of 201 preclinical articles on microvascular injury (MVI) after acute myocardial infarction, using mouse, rat, and pig ischemia-reperfusion models, assessed via the AI tool Imagetwin for inappropriate image duplication or manipulation.
57 of 201 articles (28.4%) were deemed problematic: 46 for image-related concerns and 11 for data-related concerns; 66.7% of problematic articles had corresponding authors affiliated with Chinese institutions, and the top publishers by absolute count were Elsevier (11), Springer (9), and Wiley (9).
Findings rely on a single AI-based detection tool (Imagetwin), which may miss some manipulations or flag false positives; only preclinical MVI studies using three animal models were screened, limiting generalizability; institutional affiliation patterns may reflect publication volume rather than proportional misconduct rates.
Clinicians and researchers should critically appraise preclinical MVI evidence given that roughly 1 in 4 articles in this space may be compromised. Publishers, peer reviewers, and readers should apply heightened scrutiny to image-based data in this literature.
