The surveillance state of healthcare
A combination of sensors and AI could turn our homes and hospitals into data generating-machines
“I'm sorry Dave, I'm afraid I can't do that”
-HAL 9000
The unsettling AI movie villain from ‘2001: A Spacey Odyssey’ is arguably the first example of Ambient Intelligence (in popular culture at least), and perhaps a warning for how technology could take our health into it’s own hands!
Let’s get to it…
🤷♂️ Problem
Capturing healthcare data is an active process. It requires patients, carers or health professionals to manually enter information into digital health tools.
💡 Solution
Use sensors (they’re getting cheaper) and AI (computing power is increasing) to make data collection passive
📖 Terms
Ambient Intelligence (AmI) - electronic environments that are sensitive, adaptive, and responsive to human needs, habits, gestures, and emotions
Internet of Things (IoT) - A network of physical devices – from simple sensors to smartphones and wearables – connected to each other and the internet
Sensors - A device that can detect physical property and then respond to it. Examples include colour sensors (cameras), temperature sensors (infrared), distance measuring sensors (lidar), and air pressure sensors (microphones)
HAL 9000 - The Heuristically Programmed ALgorithmic Computer is a sentient AI that controls the systems of the Discovery One spacecraft in Stanley Kubrick’s 2001 A Space Odyssey…
📚 History
As with many innovations in health, the application of AmI to healthcare started with futurists in the consumer electronic space
A team led by American entrepreneur, professor and venture capitalist Eli Zelkha coined the term ‘Ambient Intelligence’ in 1998
They prepared the keynote speech for the ‘Digital Living Room Conference’ hosted in California where the phrase was first used publicly. (If you have 30 minutes - watch this distinctly 90s coverage of the conference. Some of the predictions about computing and electronic trends are spot on).
The theme of the conference was to envision the future of consumer electronics. Check out this particular passage from the keynote:
The snapshot I’m suggesting is one in which technology anticipates our needs, in which the intelligence is ambient – much like the light in this room, satisfying our need to see without our even being conscious of it, pervades the entire room. And as long as our needs don’t change, the ambient light continues to unmediatedly satisfy the need.
Ideas around AmI and health didn’t emerge until the 2000s. This paper by Riva Guiseppe published in 2003 started to explore possible use-cases including bio-sensing, clinical data collection and therapy provision. Guiseppe correctly predicted how wireless technologies would allow us to develop natural human interfaces that would facilitate human-computer interactions through voice and gestures.
More recent developments in data science and artificial intelligence combined with the falling cost of sensors has brought the promise of AmI (as envisaged back in the late 90s) closer to reality.
💼 Use cases
Independent living. Monitor the elderly or vulnerable at home, alert family or carers when routines are missed or falls are detected
Surgical assistants. Monitor operations to detect instruments, stages of surgery, and assess surgeon skill. Surgical outcomes might improve.
Clinical note taking. Bring AmI into the consultation room to automatically record and contextualise clinical conversations. (Now there’s no excuse for poor eye contact).
Disease monitoring. Biomarkers that are amenable to sensor detection could help with chronic disease monitoring. For example, changing mobility patterns could serve as a predictor of worsening dementia symptoms allowing for early clinical intervention.
👥 Players
Consumer solutions
Amazon and Google. These guys are leveraging their existing smart speaker hardware and accompanying developer tools to provide the platform for companies to deliver healthcare solutions. Although purists might not see smart speakers as true Aml (they are not yet doing passive monitoring in the background without any human initiation - or so we’re told), they are well positioned to expand on their current offering of healthcare tools.
Quil is a digital health platform provider. Their latest product ‘Assure’, is an ambient sensor and monitoring platform to support the elderly in their homes (and provide their families with peace of mind about how Nana is getting on).
Commercial solutions
Nuance is a speech recognition company that has developed ‘Ambient Clinical Intelligence’ - an EHR adjunct that can capture and contextualise patient-clinician interactions. (They have a scripted demo you can check out which is filmed in the weirdest consultation room we’ve ever seen).
Google was developing ‘Dr Liz’ - an AI product that sounds very similar to Nuance’s solution above. Not much has emerged from this work-stream since the concept was first announced back in 2018. (#scrapped)
Research
Stanford’s ‘Clinical Excellence Research Center’ (CERC) is headed up by founder Professor Arnold Milstein. One of their projects focuses on Aml for a variety of healthcare applications. Read their recent Nature paper on Aml here.
MIT has had a bunch of teams working in this space. ‘Project Oxygen’ is one such example with more recent publications coming out of the ‘Fluid Interfaces’ group led by Professor Pattie Maes.
🔮 Predictions
Privacy will be one of the biggest obstacles facing widespread use of AmI. Consumers will need to decide if they want sophisticated cameras and microphones in the most intimate areas of their homes.
Consumer AmI for the home won’t reduce health inequalities (at least initially). The array of sophisticated sensors and accompanying platform access will probably be accessible to those individuals or families with deep pockets.
Health insurers will likely jump on any AmI product that gives them more insight into the way people lead their lives. Much like the black boxes (telematics) added to cars to monitor driving styles, the insurance perks might be appealing, but an all-seeing eye is unsettling.
Smart speakers are fast becoming a home staple in many countries. The next generation of these devices (which might include more sophisticated sensor technology) could unleash a wave of AmI health applications.
If AmI starts to focus on clinical skill evaluation (in the operating room or elsewhere), doctors and healthcare professionals will probably push back against widespread use of the technology
🌅 Opportunities
Low tech. Canary Care is a UK company supporting elderly people at home. Their USP? Sensors without cameras or mics that monitor things like temperature, movement and door activity. Why overcomplicate it?
Usability & Co-design. The promise of automatic note taking is probably appealing to a lot of clinicians. But how that technology effectively integrates into a clinical workflow is key. Inaccurate transcribing that requires a quick proof read before being committed to the electronic health record may leave some healthcare professionals…frustrated.
Accuracy. The passive nature of AmI will only be useful if the data that is captured is of high quality. Proposed use-cases are currently limited by the accuracy of the inferences the AI makes from ingested data.
Simplified terminology. AmI, IoT, human-computer interfaces, smart homes, smart hospitals, computer vision (and more). The vocabulary is extensive, confusing and probably impenetrable for most people. The nomenclature around AmI needs to be simplified and consolidated to facilitate better public engagement.
Security. Wide-scale deployment of AmI will require the strictest security assurances. Data leaks are probably inevitable, but how they’re dealt with and mitigated will make or break public trust.
🔗 Links
This article from 2017 describing the impending AmI revolution
Catch up on the HealthTrends you’ve missed:
Know someone who’d love HealthTrends? Share it with them now…