AI model predicts heart damage from ECG in seconds, potentially avoiding costly delays
Researchers have developed a deep-learning system that predicts dangerous heart enzyme elevation directly from electrocardiograms with 88% accuracy, matching specialist judgment without waiting for blood tests. The breakthrough could reshape emergency-room protocols and reduce unnecessary hospital admissions by flagging low-risk chest-pain patients faster.
Originaltitel: Multi-modal deep-learning troponin prediction from electrocardiograms and demographic data
Aims Electrocardiograms (ECGs) and troponin (Tn) testing are essential tools for the diagnosis and management of cardiac conditions. Prompt diagnosis using these tools can significantly improve patient outcomes.Methods and results The objective of this study was to design and create a deep-learning model capable of predicting high-sensitivity troponin (hs-Tn) elevation in patients undergoing chest-pain triage. We developed a novel, multi-modal, externally validated deep-learning model that incorporates ECG data, age, and sex to predict high-sensitivity troponin-T elevation. The dataset used for this study was multi-centre and externally validated, drawing from data collected in two emergency rooms. The study population included all patients presenting to the ER with either chest pain or dyspnoea during the study period, where an ECG was recorded and a Tn test was performed. The model was trained on a dataset comprising 35 821 ECGs, with a positive fraction of 35.7%. It achieved an internal area under the receiver operating characteristic (AUROC) of 0.8958 +/- 0.0040 (95% CI) and an AUROC of 0.8765 +/- 0.0110 in external validation. The model's Score-CAM saliency maps demonstrated high activation from the ST-segment, indicating that the model draws information from relevant ECG segments.Conclusion This study presents new opportunities for enhancing triage processes, enabling more rapid and accurate alerts to physicians regarding acute myocardial infarctions. The primary benefit of predicting Tn elevation lies in the objectivity of the label compared with compounded clinical outcomes and diagnoses.