AI · SEMICONDUCTOR · FACT CHECK
SKD WIREPartly True
An AI-based electrocardiogram (AI-ECG) analysis is reported to predict the risk of atrial fibrillation in patients newly diagnosed with diabetes, with the high-risk group said to show roughly eight times higher incidence. A review of the underlying study abstract, however, found that the '8 times' figure and the sample size and incidence figures do not match the abstract, and some metrics, including the AUC score, could not be independently confirmed.
The claim, based on material published by Deep Cardio, states that AI-ECG analysis can identify newly diagnosed diabetes patients at elevated risk of atrial fibrillation, and that the high-risk group's incidence rate is eight times that of others.
The '8 times' figure cited in the claim does not match the values presented in the study abstract, and the reported sample size and incidence rates also diverge from the abstract. In addition, certain performance metrics, such as the AUC, could not be verified against available materials.
The core finding — that AI-ECG can stratify atrial fibrillation risk among newly diagnosed diabetes patients — is consistent with the reported research direction, but the specific quantitative figures in the claim remain unconfirmed in their exact form.
The image material accompanying the claim was published on Deep Cardio's website.
The central assertion that AI-ECG can predict atrial fibrillation risk in newly diagnosed diabetes patients is supported, but the headline '8 times' figure and related sample size and incidence numbers do not match the study abstract, and metrics such as the AUC remain unverifiable. The claim is therefore Partly True.
