GPU + AI = The future of health monitoring


Artificial intelligence (AI) is a hot topic. Amid all the noise about generative AI tools like ChatGPT and MidJourney, the myriad of other powerful AI applications has been drowned out. One such shift we foresee is that of health monitoring via the PC webcam. Enabled by increasingly powerful graphical processing units (GPUs) and tensor processing units (TPUs), as well as the computer vision algorithms they enable, the PC could surpass the smartwatch in this domain.

Health monitoring, once confined to expensive medical facilities, is expanding its horizons to the realm of everyday technology. PC webcams, ubiquitous in modern homes and workplaces, are emerging as transformative tools for proactive health monitoring. By harnessing the power of AI-based algorithms and modern GPUs, we are on the cusp of a new era in healthcare. Individuals can now gather more data and take control of their health like never before.

The GPU is key to the rise of AI-based health monitoring

The increasing resolution of modern cameras has given rise to the ability to monitor several biomarkers. A systematic review of vital signs monitoring using cameras found the following parameters can be monitored today with reasonably good accuracy:

  Parameter Abbreviation
Vital parameters Heart rate (or pulse) HR
Respiratory frequency (or breathing rate) fR
Blood pressure (systolic and diastolic) BP
Blood oxygen saturation SpO2
Body skin temperature BST
Derivable parameters Heart rate variability HRV
Pulse transit time PTT
Body mass index BMI


There are already consumer-facing applications of such technologies:

  • The Nuralogix Anura app uses a phone front-facing camera to capture a range of health parameters including blood biomarkers, estimates of body mass index, mental health indicators, and cardiovascular and other health risks.
  • Vastmindz Visix is a Microsoft Teams application that passively monitors vital signs. It is pitched as an addition to “corporate wellness programs”.
  • PanopticAI’s Vitals is a similar app and pitches itself as having “medical-grade accuracy”.
  • Mitsubishi Electric Research Labs HealthCam is still in the prototype stage, but measures heart rate, blood oxygen level, and body temperature.

However, these biomarkers are quite straightforward and the sorts of things any GP would measure in a standard check-up. Frequent monitoring will doubtless provide value to many, but there is a new frontier of health monitoring that goes beyond this surface level.

More work is underway to vastly expand the parameters that can be captured from a camera. And increasing computational power in GPUs will unlock ever more innovation in these fields. This is what differentiates a PC from a smartphone or smartwatch: the phenomenal parallel processing in GPUs and TPUs unconstrainted by miniaturisation. These processor-heavy AI-based models can provide passive monitoring of a much wider range of biomarkers, all being captured in the background.

Detection of health conditions will no longer come from a loved one noticing an unusual eye twitch or someone finally visiting the doctor after months of intensifying chest pain. We will get a notification well in advance and address emerging issues far more rapidly than today’s technology allows.

Blepharometry will be a large part of this transformation

Blepharometry is the study of eyelid movements. Interestingly, the eyelid is controlled by two muscles in the brain. The pons stimulates the orbicularis oculi muscle, which closes the eyelid. The midbrain stimulates the levator palpebrae superioris muscle, which opens the eyelid.

Tracking a person’s eyelid signal offers a window into their cognitive state across many domains. As the brain is subjected to various forms of impairment, such as drowsiness, intoxication, a neurodegenerative condition, or lack of oxygen, the coordination between these two muscles deteriorates. This can be detected in the eyelid signal, and cameras in consumer-grade devices have reached a tipping point at which our team at Optalert can now derive a signal of high enough quality for our industrial-grade algorithms.

We first used blepharometry to quantify impairment from drowsiness using the Johns Drowsiness Scale (JDS). We successfully commercialised this technology in mining and transport, as well as the automotive industry. Our R&D team has now turned its attention to a range of health conditions with astounding results.

Obstructive sleep apnoea (OSA) is one such condition. It affects 13.2% of people aged 30-69 worldwide, yet 50-80% of cases go undiagnosed. For those who go untreated, they live their lives with excessive daytime sleepiness and an invisible weight holding them back. At Optalert, we have developed an algorithm that detects OSA with 84% accuracy. It only uses one input: a subject’s eyelid signal.

We also have active research focussed on early detection of Alzheimer’s, Parkinson’s, and other neurodegenerative conditions, all with promising results. Detecting these conditions years in advance of when they are visibly symptomatic could enable an intervention that adds years or even decades to people’s lives.

Imagine how many lives we could improve by passively monitoring people’s eyelids as they worked. This is the bright future we envision. This is what drives us.

Health monitoring via webcam and GPU offers many benefits

  • Privacy and security: GPUs enhance data privacy and security. By enabling on-device processing, sensitive health information can be analysed locally without the need for data to leave the user’s device.
  • Convenience and accessibility: Webcams are already an integral part of our digital lives. Leveraging these devices for health monitoring eliminates the need for specialised equipment, making it convenient and accessible for a broad population. Individuals can monitor their health from the comfort of their homes or workplaces, promoting regular check-ins and early intervention.
  • Non-intrusive monitoring: Traditional health monitoring methods often require individuals to wear devices or undergo specific tests. Webcam-based monitoring is non-intrusive, allowing individuals to carry on with their daily routines while their health data is collected discreetly.
  • Real-time insights: AI algorithms can analyse webcam data in real-time, providing immediate feedback on vital health metrics. This rapid analysis enables individuals to track changes and respond promptly to any abnormalities.
  • Cost-effectiveness: By utilising existing webcams, the cost of implementing health monitoring is significantly reduced. This democratisation of healthcare empowers individuals regardless of their socioeconomic status.

As with all health technologies, there are challenges

  • Data quality and reliability: Ensuring accurate and reliable data capture from webcams is paramount. Factors such as lighting conditions and camera quality can impact the quality of video data, potentially affecting the accuracy of health assessments.
  • Algorithm accuracy: The accuracy of AI algorithms is crucial for reliable health insights. Rigorous training and validation are necessary to fine-tune algorithms to provide accurate results across diverse populations. They also need to be validated against the ground truth of current measurement devices.
  • Ethical and regulatory considerations: The collection and analysis of health data raise ethical concerns related to privacy and consent. Regulatory frameworks must be developed to safeguard user data and ensure compliance with healthcare standards.
  • Integration with healthcare ecosystems: Integrating webcam health monitoring into existing healthcare systems and Electronic Health Record (EHR) platforms requires seamless interoperability. Collaboration between technology developers and healthcare providers is vital.
  • User acceptance: Convincing users of the reliability and benefits of webcam health monitoring is essential. Educating individuals about the capabilities and potential of these technologies can drive adoption. Further, people will need to trust that their data will only be used in ways they permit via opt-in consent and ability to revoke at any time.

In spite of the challenges, we see a bright future in this space

The future of health monitoring is being rewritten through the convergence of AI-based algorithms, high-resolution PC webcams, and modern GPUs. This triumvirate of technology empowers individuals to actively participate in their health management, transforming the healthcare landscape from reactive to proactive. The convenience, accessibility, and real-time insights offered by webcam health monitoring hold the potential to revolutionise how we approach preventative healthcare.

As we navigate challenges related to data accuracy, privacy, and regulatory compliance, collaboration between technology innovators, medical professionals, and policymakers will be critical. The journey to widespread adoption of health monitoring via the PC webcam is exciting, with the promise of healthier lives, improved patient outcomes, and a healthcare system that is more connected and patient-centric than ever before. As the pixels on our screens translate into a clearer picture of our health, the future of health monitoring through PC webcams is poised to bring us closer to a healthier and more empowered society.

Want to learn more about Optalert’s technology?

Please contact us if you want more details on our drowsiness detection, OSA breakthrough, or other technologies.