MinderCare is an NHS-linked home-monitoring service designed to help clinicians spot changes in the health and daily routines of people living with dementia. Developed by the UK Dementia Research Institute (UK DRI) with Imperial College Healthcare NHS Trust, it combines discreet sensors, artificial intelligence and daily clinical review to build a picture of what is happening between appointments. The service launched in North West London in 2025 after more than a decade of research into technology-assisted dementia care.
The technology does not depend on cameras or microphones. Instead, MinderCare uses door, movement and appliance sensors together with a sensor mat beneath the mattress. These devices can capture changes in activity, sleep, breathing and heart rate. AI-enabled algorithms analyse the resulting patterns and generate risk notifications, while nurses and specialist clinicians review the information and decide whether a person or their care plan needs attention.
What MinderCare actually monitors
MinderCare is built around a relatively simple idea: establish what is normal for an individual and then look for meaningful departures from that pattern.
That is different from a conventional medical device that records a single measurement at a particular moment. Someone’s movement around the house, sleep behaviour or nighttime activity can vary considerably from day to day. The potential value of continuous monitoring comes from seeing those measurements over time and combining them.
The current system uses several kinds of sensors:
| Sensor or data source | What it can monitor | Potential clinical significance |
|---|---|---|
| Movement sensors | Activity and movement around the home | Changes may indicate reduced activity, altered routines or increasing care needs |
| Door sensors | Use of external doors and unusual nighttime exits | Can highlight changes in nighttime behaviour or potential safety concerns |
| Appliance sensors | Patterns associated with everyday household activities | Changes can provide information about alterations in established routines |
| Bed sensor/mat | Sleep, movement, breathing and heart rate | Can help clinicians track sleep disruption and physiological changes |
| Combined sensor data | Patterns across multiple measurements | AI analysis can flag changes for clinical review |
Imperial says the devices do not record sound or video. The intention is to gather useful information while avoiding the continuous visual surveillance associated with cameras in the home.
Where the AI comes in
The sensors themselves are not the most important part of MinderCare. The harder problem is interpreting the large amount of data they generate.
Minder uses AI and data-analysis techniques trained to identify clinically important events and provide alerts to a monitoring team. The system can look across environmental, activity and physiological information rather than treating every measurement independently.
This creates a three-stage process:
Sensing → analysis → clinical response
The sensors collect information in the home. Algorithms analyse patterns and generate risk notifications. The clinical team then reviews those alerts and the wider context before deciding whether action is appropriate.
That last stage is crucial. An AI-generated notification is not itself a diagnosis.
A change in breathing rate, for example, might have several explanations. Likewise, someone leaving their home at an unusual time does not necessarily mean they are experiencing a medical problem. The system is intended to give clinicians an additional source of evidence rather than make autonomous medical decisions.
Imperial describes the data as being reviewed daily by the MinderCare monitoring team, which can then coordinate with patients, carers, GPs and community services where necessary.
How the system could spot an infection
One of the more significant applications is the detection of possible changes associated with infection.
Respiratory and urinary infections can cause serious complications in older people, while symptoms may be difficult to recognise in someone living with dementia. MinderCare’s approach is to look for combinations of subtle physiological and behavioural changes rather than waiting for an obvious crisis.
The system can monitor measurements including breathing and heart rate alongside activity patterns. A sustained change can trigger an alert for clinical review.
According to Imperial’s March 2026 update, the infection-risk algorithm had identified more than 165 clinical events since the programme began. A separate nighttime-safety algorithm had identified another 85 episodes of nighttime wandering.
Those numbers demonstrate that the system is producing clinically relevant alerts. They do not, on their own, demonstrate that the technology prevents hospital admissions, reduces mortality or improves long-term health outcomes. Those questions require broader evaluation.
Sleep is another major use case
Sleep disruption is common among people living with dementia and can affect both the individual and their caregiver.
The bed sensor provides a way to measure changes in nighttime activity while also capturing breathing and heart-rate information. Clinicians can use those measurements to see whether a person’s sleep pattern changes following an intervention or adjustment to their care.
This could make a difference to how clinicians assess problems that would otherwise rely heavily on retrospective accounts.
A caregiver might know that someone has been sleeping badly, for example, but continuous measurements can provide a more detailed picture of when the disturbance occurs and how the pattern changes over time.
The technology therefore shifts part of dementia care from occasional observation toward longitudinal monitoring.
What happens when the AI raises an alert?
MinderCare is not designed as a fully autonomous system.
Alerts and data are reviewed by a dedicated clinical team that includes dementia specialists and nurses. The team can examine the person’s activity over time and determine whether the change warrants intervention.
Depending on the circumstances, that intervention could involve advice to a patient or caregiver, changes to a care plan, communication with another healthcare professional or further clinical assessment.
This human-in-the-loop model matters because behavioural data is inherently contextual.
A person might move less because they are unwell, but they might also move less because a family member is visiting. Someone might leave the house during the night because of confusion, but an unusual nighttime exit could also have an entirely ordinary explanation.
AI can identify patterns at a scale that would be difficult for a person to monitor continuously. It cannot automatically understand every reason behind those patterns.
Privacy is part of the engineering challenge
Monitoring someone inside their own home creates an obvious privacy problem.
MinderCare’s design attempts to reduce that concern by using sensors that do not capture conventional audio or video. UK DRI describes the platform as using privacy-preserving data collection, with limited information shared centrally.
That does not mean there are no privacy implications.
Movement patterns, sleep behaviour, physiological measurements and household routines can reveal highly sensitive information even without a camera. Over time, such data can create a detailed picture of a person’s daily life.
The relevant questions therefore include who can access the information, how long it is retained, how it is secured, what is shared with other organisations and how patients and carers can exercise control over their participation.
MinderCare’s research participants consent to take part in an HRA-approved research study, and the programme is also being used to develop and validate digital biomarkers and assess healthcare and caregiver impacts.
The privacy advantage of avoiding video is meaningful, but it should not be confused with privacy being solved.
The technology is becoming more sophisticated
The sensor network is not necessarily the final form of MinderCare.
A current 2026–2027 project at Imperial College London is testing radar technology within the MinderCare environment. The project aims to validate a novel radar sensor in real-world dementia care, adding another way of monitoring movement and activity without requiring conventional cameras.
This is part of a broader direction in ambient healthcare technology: replacing intrusive observation with increasingly capable sensors that can operate continuously in the background.
The same technological shift is appearing in other AI systems that move from simply responding to users toward continuously sensing and acting within their environment. The security implications of that transition are explored in the evolution of AI agents into internet actors.
For dementia care, however, the objective is not to turn the home into an autonomous AI environment. It is to provide clinicians with information that would otherwise be difficult or impossible to collect.
MinderCare is also a research platform
The distinction between a clinical service and a research project is important.
MinderCare was developed through the UK DRI’s Care Research & Technology Centre and is being used within an ongoing research programme. UK DRI says the wider Minder platform has been trialled in around 100 homes and combines environmental, physiological and activity data with AI analysis and clinical monitoring.
The programme is also examining whether digitally enabled care produces benefits beyond individual clinical alerts.
Researchers are studying healthcare use, digital biomarkers and the economic effects of the service, including possible impacts on caregiver wellbeing. Imperial’s March 2026 update describes plans for further testing and expansion, including work using the Discover-NOW dataset.
That research matters because a technically impressive monitoring system is not automatically a useful healthcare system.
For NHS adoption at scale, the technology would need to demonstrate that the information it produces leads to worthwhile clinical outcomes and that those benefits justify the costs, infrastructure and workload associated with monitoring it.
What MinderCare can—and cannot—tell clinicians
The most useful way to understand the system is as an additional layer of evidence.
It can potentially show that a person’s normal pattern has changed. It can help clinicians identify changes in sleep, movement, nighttime activity or physiological measurements. It can provide an objective record that may complement information from patients and caregivers.
But it cannot establish the cause of every change.
| MinderCare can potentially provide | It does not automatically provide |
|---|---|
| Continuous information about activity and routines | A dementia diagnosis |
| Changes in sleep and nighttime behaviour | A definitive explanation for why sleep changed |
| Breathing and heart-rate trends | A medical diagnosis from one abnormal reading |
| Alerts about unusual patterns | Proof that an emergency is occurring |
| Longitudinal information for clinicians | A substitute for clinical assessment |
| Additional evidence for care decisions | Guaranteed prevention of hospitalisation |
That distinction is particularly important when AI is involved. A prediction or risk notification describes what an algorithm has identified in the data; it does not necessarily establish what is happening medically.
From research project to NHS care
The significance of MinderCare is that the underlying technology has moved beyond a laboratory demonstration.
Imperial College Healthcare NHS Trust launched the NHS service in 2025, integrating it with the North West London Virtual Hospital. The service is supported by specialist clinical staff and draws on technology developed through the UK DRI research programme.
That creates an unusual model in which research, monitoring technology and clinical care operate together.
The potential advantage is feedback in both directions. Clinicians receive new information about patients living at home, while researchers can study how the technology performs under real-world conditions.
But the programme is still at a stage where its broader effectiveness and scalability are being evaluated. UK DRI has described the wider goal as enabling people with dementia to remain independent at home for longer while supporting more personalised care.
The evidence needs to establish how consistently that goal is achieved outside individual examples.
The bigger question is what happens after the alert
MinderCare demonstrates an important principle in healthcare AI: detecting a problem is only the first half of the equation.
An algorithm may identify an unusual pattern with impressive speed. The practical value depends on what happens next. Is a clinician available to review it? Can the patient receive an assessment quickly? Is appropriate community care available? Does acting on the information actually improve the outcome?
Those questions will determine whether home AI monitoring becomes a significant part of dementia care or remains primarily a promising research technology.
The most important development to watch is therefore not simply whether MinderCare adds more sensors or produces more sophisticated algorithms. It is whether longer-term clinical and economic evidence shows that continuous, privacy-conscious monitoring leads to earlier and more appropriate care without creating an unmanageable stream of false alarms or surveillance concerns.
If that evidence emerges, the technology’s importance will extend beyond dementia. It would demonstrate a broader model of healthcare in which clinicians can monitor changes in a person’s everyday environment and intervene before a relatively subtle problem becomes a crisis. For now, MinderCare offers a working example of that model—but the evidence for its ultimate impact is still being built.