Log in to MyACC
Smartphone-Based Recognition of Heart Failure
Quick Takes
- Seismocardiography is a noninvasive technique, which measures cardiac-induced mechanical vibration and includes accelerometers and gyroscopes.
- Smartphone-based seismocardiography using microelectromechanical sensors was able to reliably acquire signals and differentiate patients with HF from control subjects.
Study Questions:
Can cardiac motion measured using smartphone-based seismocardiography identify patients with heart failure (HF)?
Methods:
This was a prospective observational study, which examined cardiac motion signals acquired from a commercially available smartphone. Patients with and without HF were enrolled in both the inpatient and outpatient settings from three centers (two in Finland and one in the United States). Patients with severe valvular disease, prosthetic valves, and those paced >1% of the time were excluded. Cardiac signal acquisition was carried out using a Samsung phone, which was placed over participants’ sternums for 2 minutes while supine and when at an incline. Accelerometer (linear acceleration) and gyroscope (angular velocities) signals were uploaded from the smartphone to a secure web-based platform and features extracted (signal amplitude, area, time intervals). After data acquisition and signal processing, data were aggregated and features input into a machine learning classifier. Bootstrap aggregation was used on derivation (excluding patients with atrial fibrillation [AF]) and validation (including patients with AF) datasets. A decision threshold of 0.5 was used to classify HF cases.
Results:
The study enrolled 1,025 participants with a final analytic set of 1,003 participants after exclusions, 217 of whom had HF (174 inpatients and 172 outpatients). All patients had American College of Cardiology/American Heart Association (ACC/AHA) stage C HF. In the inpatient setting, 74% had HF with reduced ejection fraction (HFrEF) and 74% New York Heart Association (NYHA) class III/IV HF. In the outpatient setting, 58% had HFrEF and 15% NYHA class III/IV HF. The diagnostic performance of the algorithm for the identification of HF included an area under the receiver operating characteristic curve (AUC) of 0.95, sensitivity of 0.85, specificity of 0.90, diagnostic accuracy of 0.89, positive likelihood ratio of 8.5, and negative likelihood ratio of 0.17. Model performance was similar in both the inpatient and outpatient settings and in subgroup analyses based on age, sex, body mass index, and presence of AF. In the inpatient setting, sensitivity for HF with preserved EF (HFpEF) was 0.76 and for HFrEF 0.89. In stable outpatients, sensitivity for HFpEF was 0.81 and for HFrEF 0.88. AUCs were significantly higher for gyroscope as compared with accelerometer signals. S3 angular velocity features best identified HF patients.
Conclusions:
The authors conclude that smartphone-based seismocardiography is feasible and can be used to identify patients with HF.
Perspective:
Mobile health technology for point-of-care diagnostic testing for HF offers the potential to enable earlier diagnosis and democratize access to health care. A prior study of single-lead electrocardiograms from a wearable device was able to accurately diagnose left ventricular systolic dysfunction (Khunte A, et al., npj Digit Med 2023;6:124). Strengths of the current approach include reliance on a smartphone, which can enable point-of-care testing broadly given high rates of smartphone ownership worldwide, and superior diagnostic performance irrespective of EF or clinical setting.
A number of important limitations are of note, however, before this tool can be widely implemented. First, signal acquisition was obtained in conjunction with the investigative team. Whether signal acquisition could be undertaken by patients independently within their homes, the environment in which this technology has the greatest potential, is still unknown. Second, the control group was not well characterized and patients with some common forms of structural heart disease were excluded from the study (e.g., prosthetic valves, paced rhythms). The performance of the model amongst patients with structural heart disease but without HF is an important and unanswered question. Third, downstream protocols would be required to ensure appropriate testing and treatment for patients with new HF. Finally, the study provides no information on race/ethnicity, and studies are needed to demonstrate the performance of this tool amongst racially and ethnically diverse individuals.
Clinical Topics: Heart Failure and Cardiomyopathies, Acute Heart Failure
Keywords: Heart Failure, Smartphone
< Back to Listings
