FaceHeart Vitals Software Development Kit (FH vitals SDK)

The FaceHeart Vitals Software Development Kit (FH vitals SDK) aims to provide non-invasive pulse rate spot measurements by analyzing facial video streams captured by consumer-grade cameras. Its methodology relies on optical camera-based detection of subtle facial skin color variations indicative of blood volume changes, which are then processed using face recognition and signal processing algorithms to compute the pulse rate for stationary adult users.

The FaceHeart Vitals Software Development Kit (FH vitals SDK) is a software-only product designed for non-invasive spot measurements of pulse rate in adults (18 years and older) who are not in critical care and do not require continuous monitoring. It functions by analyzing a facial video stream captured by standard cameras on commercial mobile devices, laptops, or computers running Android, iOS, or Windows. The system utilizes face recognition to detect the user’s face, then processes the continuous face image data through proprietary signal processing and algorithms to compute and display the pulse rate in real-time. The SDK is intended for use when the subject and the camera are stationary, and it supports minor movements like speaking or slight nodding, operating optimally with illumination above 300 Lux and a measuring distance of 0.5-1.5 meters. Clinical performance data from a study of 527 participants demonstrated that, compared to a Philips MX100 ECG, the FH vitals SDK achieved a pulse rate deviation within ±3 bpm for approximately 93.8% of subjects on average across various tested devices, over a measurement window of 1 minute, and within a pulse rate range of 50-180 bpm. Non-clinical testing confirmed device compatibility (e.g., laptops with C920, iPhone 13 series, Samsung S22 series), characterized system performance metrics (memory, CPU usage), and highlighted that significant motion can considerably degrade accuracy, with mean absolute errors ranging from 11.77 bpm to 42.06 bpm. The SDK, which tracks only one face and provides various warning messages (e.g., for missing face, bad position, motion), is not intended for patients with known heart arrhythmias or to treat patients; its results should complement professional medical care. Device manufacturers integrating this SDK are responsible for the final medical device's regulatory approval and ensuring its suitability for the intended use.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the algorithm differentiate between actual pulse-related color changes in the facial video stream and noise introduced by minor head movements, changes in ambient lighting, or varying skin tones beyond the Fitzpatrick scale types tested?; What specific signal processing techniques are employed to extract the pulse rate from the facial video data, and how are these optimized for different camera hardware (e.g., C920, iPhone, Samsung) and operating systems (Windows, iOS, Android)?; Given the significant impact of motion (error up to 42.06 bpm), what is the precise threshold for "minor movement (speak or nod slightly)" beyond which the "Motion" warning is triggered and measurement accuracy is compromised, and how was this threshold validated?; The SDK is not for patients with known heart arrhythmias; how does the system perform or alert users if an arrhythmia is unknowingly present during measurement, and could it misinterpret arrhythmic patterns?; What were the specific selection criteria for the 527 participants in the clinical study, particularly regarding the distribution of cardiovascular disease histories (arrhythmia, CAD, HF, hypertension), and how did these conditions potentially influence the ±3bpm accuracy reported?