properties (compliance, stiffness, radius) that cannot be observed directly.
This project will develop a physics-based computational model of contactless BP measurement by coupling a haemodynamic model of the facial arterial network with PanopticAI's rPPG technology. Working with PanopticAI (a leading contactless vital sign monitoring company) and King's cardiovascular modelling experts, the student will quantify which vascular parameters are recoverable from facial video, establish the theoretical limits of calibration-free BP estimation, and develop physics-grounded calibration strategies, delivering the evidence needed for regulatory clearance and clinical trust.
Project description
Blood pressure (BP) is the most important modifiable risk factor for cardiovascular disease, the leading cause of death worldwide. Yet BP is still measured using an inflatable cuff: intermittent, uncomfortable, and unable to capture the continuous ambulatory patterns that predict risk. Remote photoplethysmography (rPPG) extracts the pulse waveform from minute skin-colour changes captured by an ordinary RGB camera, and can in principle estimate BP contactlessly, continuously, and at scale. Despite strong academic and industrial progress, no group has achieved calibration-free, cuffless BP clearance from any major regulator. The remaining barrier is fundamentally physical, not merely algorithmic.
The facial rPPG signal encodes the arterial pulse wave. Features such as pulse transit time (PTT) between facial regions and waveform morphology relate to pulse wave velocity (PWV), which depends on arterial stiffness through the Moens-Korteweg relation. Converting PWV to absolute BP, however, requires knowledge of vessel wall elasticity, thickness, and radius, and these individual parameters cannot be separately identified from optical signals alone. Only composite quantities (such as the product of elasticity, wall thickness, and inverse radius) are observable. This creates a per-subject ambiguity that is difficult to resolve from optical signals, with vascular tone and absolute BP becoming entangled. These identifiability limits are poorly characterised for facial rPPG and currently obstruct both clinical trust and regulatory clearance.
This project addresses the problem from first principles by developing a physics-based digital twin of contactless BP measurement. The student will develop a forward haemodynamic model of the facial and cervical arterial network: given vessel properties and BP, the model will predict the pulse wave and the optical signal a camera would observe. The model will be coupled with PanopticAI's contactless rPPG technology, used as a reference and validation platform, to interrogate the full measurement chain: which arterial parameters can be recovered from facial video, where the calibration ambiguity lies, and how it can be reduced using observable proxies such as perfusion amplitude, heart-rate variability, and waveform shape. The work will build the physics and calibration evidence that contactless BP needs to move from controlled validation to trustworthy, generalisable clinical use.
The project is co-sponsored by PanopticAI Limited, a Hong Kong deep-tech company and a leader in contactless vital sign monitoring, with an active contactless BP programme and established regulatory experience in contactless vital sign technologies. PanopticAI provides the industrial supervisor (Nick Chin, CTO/Co-Founder), its rPPG technology pipeline, existing ethically approved clinical datasets, and regulatory context. The ideal candidate will have strengths in at least two of the following areas: biomedical engineering, applied mathematics or physics, cardiovascular modelling, signal processing, or machine learning, with strong computational skills (numerical methods, Python or MATLAB) and an interest in inverse problems. Expected outcomes include a validated forward haemodynamic model, a formal identifiability and calibration analysis for contactless BP, publications, and invention disclosures, delivering the scientific evidence base for the first calibration-free contactless BP clearance.
