The study, published in PLOS Digital Health and authored by researchers at Aalborg University Hospital, Aarhus University Hospital and Odense University Hospital, found the platform’s machine-learning algorithm achieved 90.9% sensitivity and 87.5% negative predictive value for detecting heart failure of any type, in a cohort of 218 patients. For the reduced-ejection-fraction subtype specifically, diagnostic accuracy was similar to the NT-proBNP blood test (92.9% vs. 94.7% AUC), but with markedly higher specificity, meaning far fewer false positives.

The heart-failure algorithm is investigational: it is built into Acarix’s Gen 2 device but is not yet part of the CADScor System’s CE mark or FDA clearance, both of which currently cover only the rule-out of coronary artery disease. The authors describe the results as exploratory, noting the study was conducted at just two centres in a relatively small population, and that the algorithm will require external validation in larger, independent cohorts before Acarix can pursue regulatory submission.

“Heart failure and coronary artery disease are the two biggest reasons patients end up in front of a cardiologist, and this new research opens up a whole new frontier for our CADScor System. We built CADScor to give clinicians a fast, accurate, first-line tool for cardiac triage, and these results show we can bring that same value to heart failure,” said Aamir Mahmood, CEO of Acarix.