![heart](http://media.assettype.com/healthday-en%2F2024-06%2F4933721d-9587-4ea6-a4d4-8558785f34f0%2FAdobeStock_239759115.jpeg?w=480&auto=format%2Ccompress&fit=max)
WEDNESDAY, June 5, 2024 (HealthDay News) -- A noncontact facial infrared thermography (IRT) machine learning model can predict coronary artery disease (CAD), according to a study published online June 3 in BMJ Health & Care Informatics.
Minhui Kung, from Tsinghua University in Beijing, and colleagues assessed the feasibility of using facial IRT temperature information with machine learning for prediction of CAD among individuals referred for invasive coronary angiography or coronary computed tomography angiography. A deep learning IRT image model was developed and validated for detecting CAD using facial IRT images captured before confirmatory CAD examinations. Performance of the IRT image model was compared to that of the guideline-recommended pretest probability (PTP) model on the area under the curve (AUC). To further validate the predictive value of IRT information, interpretable IRT tabular features were extracted from IRT images.
Data were included for 460 eligible participants. The researchers found that the IRT model demonstrated outstanding performance compared with the PTP models (AUC, 0.804 versus 0.713). The predictive value of IRT information was further validated by a consistent level of superior performance (AUC, 0.796), achieved with comprehensive interpretable IRT features. A satisfactory performance was upheld even with only traditional temperature features (AUC, 0.786).
"Our developed IRT prediction models, based on advanced machine learning technology, have exhibited promising potential compared with the current conventional clinical tools," the authors write.