PM1035: Rapid Non-invasive Screening for Left Ventricular Dysfunction in Children Using a Handheld Smartphone-based Device

Background: While physicians and parents of children with heart disease are taught to recognize symptoms of heart failure, many of these symptoms (e.g., tachypnea) are nonspecific. In resource limited health care settings and in the home, a need exists for a simple, non-invasive, and inexpensive method of screening for left ventricular (LV) dysfunction. Previous studies in adults have shown that a handheld smartphone-based device (Vivio) can easily and quickly derive LV ejection fraction (LVEF) from a carotid artery pulse via novel parameters called “intrinsic frequencies.” The goal of this study was to evaluate whether intrinsic frequencies measured by the Vivio device could be used to distinguish between normal and abnormal LVEF in children and young adults with normal cardiac anatomy.

Methods: The study included patients ages 0 to 21 years with normal cardiac anatomy who were evaluated by either echocardiogram or cardiac magnetic resonance imaging (MRI) for clinical indications. The only exclusion criterion was known vascular disease that would interfere with obtaining a carotid artery pulse. LVEF was calculated using Simpson’s method on echocardiogram apical images or MRI short-axis images. Carotid artery waveforms were recorded with Vivio on the same day as echocardiogram or MRI. Machine learning decision tree analysis was used to determine the optimal combination of intrinsic frequency parameters to discriminate between normal and abnormal LVEF ( < 50%).

Results: Fifty patients were included. Median age was 9.9 years (IQR 5.1-14.5 years) and 28/50 (56%) were female. LVEF was measured by echocardiogram in 46/50 (92%) and by MRI in 4/50 (8%). LVEF was abnormal in 16/50 (32%). The machine learning algorithm evaluated several classification schemes using a combination of intrinsic frequency parameters, age, and heart rate (see Table). A classification scheme based on the intrinsic frequency parameter, ω1, and this parameter adjusted for heart rate, ωi1, was able to detect an abnormal LVEF with a sensitivity of 100%, specificity of 89%, and accuracy of 92%. Consistent with prior studies of intrinsic frequencies in adults with abnormal LVEF, most children with abnormal LVEF had elevated ω1 (see Figure). Hence, ω1 played a dominant role for detection of abnormal LVEF in all classification schemes.

Conclusion: Intrinsic frequencies measured using a handheld smartphone-based device (Vivio) can differentiate between normal and abnormal LVEF in children with normal cardiac anatomy. The proposed method can be done quickly, non-invasively, and easily, allowing for its use in multiple settings by both medical professionals and family members. In the future, more diagnostic and prognostic information can be obtained as this method is refined to provide accurate LVEF quantification across all ages.