Breakthrough AI Technology Accurately Assesses Heart Failure Severity

Heart failure (HF) is a complex condition where the heart cannot effectively pump blood to meet the body’s needs due to underlying medical issues. It is marked by recurring episodes and frequent hospitalizations. HF impacts over 64 million individuals globally and remains a leading cause of death. Early identification of HF is essential for effective treatment and better patient outcomes. Traditionally, HF diagnosis involves patient interviews, electrocardiogram (ECG), echocardiogram, and lab tests, a process that can be time-intensive and requires specialized expertise. Now a new artificial intelligence (AI)-based technology has been developed to automate heart failure detection, enabling clinicians to make quicker and more informed decisions.

Simplex Quantum (Shibuya, Japan), in partnership with the University of Tokyo Hospital (Tokyo, Japan), has launched NIHA-HF, an innovative AI system designed to precisely assess the severity of HF and monitor disease progression using single-lead ECG data from widely available devices such as the Apple Watch. A key feature of this technology is the HF-index, an AI-generated metric that allows for on-demand monitoring of HF severity at home. This system holds great promise in reducing hospitalizations by facilitating earlier clinical intervention. The AI model was trained on more than 11,000 ECGs annotated by physicians.

NIHA-HF detects heart failure by analyzing a 30-second lead I ECG. Utilizing advanced signal processing combined with deep learning, this technology can be applied to both conventional 12-lead ECG devices and wearable ECG products like lead I. In a study published in the International Journal of Cardiology, the AI model, trained with ECG data from 9,518 individuals, achieved 91.6% accuracy in classifying HF severity into healthy, NYHA I–II (mild), and NYHA III–IV (moderate to severe) categories. The model showed area under the curve (AUC) values reaching 0.993, with sensitivity and specificity ranging from 89% to 97%. The HF-index was found to strongly correlate with plasma B-type Natriuretic Peptide (BNP) levels (R = 0.74), a well-established biomarker for heart failure.

"This AI model may support on-demand, non-invasive heart failure monitoring, even before symptoms appear," said Dr. Katsuhito Fujiu, Project Professor at the University of Tokyo and senior author of the study. "It could fundamentally change how we care for patients at risk of worsening heart failure by empowering both clinicians and patients with timely insights."

"These results validate the clinical foundation of our solution and show how AI can turn familiar devices into powerful medical tools," added Ryu Saito, CEO of Simplex Quantum. "We are proud to collaborate with top-tier institutions to bring this technology into everyday practice."

Related Links:
Simplex Quantum
University of Tokyo Hospital