HeartFM

The challenge

Acute decompensation in heart failure is often recognised late, resulting in emergency hospital admissions and limited opportunities for early intervention. Patients may experience gradual fluid accumulation, worsening symptoms, or physiological changes over days or weeks before reaching a crisis point requiring hospitalisation. These early warning signs frequently go undetected between scheduled clinic visits, as traditional monitoring relies on episodic assessments and patient self-reporting of symptoms. Late recognition of decompensation not only compromises patient outcomes but also drives healthcare costs through avoidable emergency admissions and prolonged hospital stays. There is a clear need for non-invasive, continuous monitoring approaches that can identify early signs of deterioration outside the clinical setting, enabling timely intervention before patients require emergency care.

The project: HeartFM

HeartFM focuses on developing and validating data-driven algorithms to detect unstable heart failure using wearable sensor data and clinical information. The project brings together clinical expertise, algorithm development capabilities and wearable technology innovation to address a critical gap in heart failure management.

TNO leads the algorithm development and validation, working with implantable cardioverter-defibrillator (ICD) data provided by Leiden University Medical Center. ICD devices continuously record physiological parameters in heart failure patients, providing a rich source of validated clinical data for algorithm training. TNO's role encompasses designing the analytical framework, developing machine learning models and rigorously validating their performance against clinical outcomes. Meanwhile, AIKON Health develops the wearable sensor patch that enables continuous, non-invasive data collection in daily life, translating the algorithm from an ICD-based development platform to a practical monitoring tool suitable for broader patient populations.

The project has demonstrated the technical feasibility of detecting unstable heart failure, with a first version of the algorithm successfully trained on the ICD dataset. This proof-of-concept establishes that machine learning approaches can identify patterns in physiological data that precede clinical decompensation. The algorithm incorporates multiple data streams to capture the complex, multifactorial nature of heart failure progression. Ongoing work focuses on refining the algorithm to improve sensitivity and specificity, reducing false alarms whilst maintaining high detection rates, and validating performance in prospective clinical studies with patients wearing the AIKON Health sensor patch.

Looking to the future

Following clinical validation, HeartFM aims to integrate the decompensation detection algorithm into routine heart failure care pathways. Remote monitoring using the wearable sensor patch could enable proactive management, with clinicians receiving alerts when patients show early signs of deterioration, allowing for timely medication adjustments or clinical review before hospitalisation becomes necessary.

Future development will explore personalised thresholds that account for individual patient baselines and disease trajectories, potentially improving detection accuracy. Widespread implementation will require collaboration with cardiology departments, healthcare systems and reimbursement bodies to demonstrate clinical effectiveness and cost-effectiveness, ultimately reducing hospital admissions and improving quality of life for heart failure patients.