Technology to help dementia carers

[Jorjix]

Hello all,

My name is Georgiana, I am a computer scientist and, together with a bright neurologist from UCLH, I'm am investigating the possibility of creating a system that automatically analyses CCTV footage in order to identify cognitive decline, fall risks or dangerous behaviours in people that suffer from dementia.

CCTVs in homes are very controversial, but would a smart system that notifies carers or family about suspicious behaviours be useful? This could be as simple as forgetting the front door open or leaving the house without notifying the carer, or more complex, such as detecting physical and mental decline, or even predict patient vulnerability which leads to falls.

I would be interested to know more about your views on such a monitoring system. Would it be of any use? Do you already use something like this? Any input would help, really.

 

[nitram]

"I would be interested to know more about your views on such a monitoring system. "

My first question would be to ask if you had any ball park figures for installation and operating costs of the system.

Are you considering a install, rent as long as required, remove system?

 

[Jorjix]

My first question would be to ask if you had any ball park figures for installation and operating costs of the system.

Are you considering a install, rent as long as required, remove system?

To be perfectly honest, we don't have a price for the service.
This is a research project and we're currently building the technology for assessing behavioural patterns from home sensors (camera is one of them). We want to test the system with people who already have cameras installed and would be interested to automatically spot anomalies when they happen, without having to constantly check the videos.

Initially, we are not considering installing the cameras, but can help with that. Our goal is to provide the behavioural analysis as a subscription base service and are wondering whether this system can help dementia carers.

 

[Jorjix]

Thanks so much for your insights; they are extremely valuable!

Everyone is so different and what may be considered decline may only be a urine infection or a shock and their behaviour changes dramatically though their illness has remained the same.

Indeed. Our assumption is that there are relevant behavioural cues that indicate different vulnerability levels and, by analysing the patients day-to-day, we can create a signature of how risk looks like for each one of them. To give you a concrete example, the way a person walks can give interesting indications if analysed over time and ideally, the system could distinguish between a temporal shock and a cognitive decline.

Having said that, it's true this is an ambitious attempt and we're first trying to assess whether such an analysis is of any use for carers or not. Again, the system could alert when forgetting the gas on, leaving the door open, or leaving the house in the middle of the night...

The physical and mental decline - that would be more invasive (bedroom/bathroom - is when a lot of things may/could occur). In a bedroom - them getting dressed or having difficulty of getting dressed. Again, bathroom remembering to wash their hands. To me these two areas are far too personal and invasive.

That's a fair point and it's something we're trying to solve, without infringing anyone's privacy. We're currently considering ways to anonymise the recorded data. Would it be easier to decide on installing a camera in the bedroom if the video was, for example, blurred?

Thanks again for your inputs.

 

[Lawson58]

I think you need to spend some time with your target clients in order to have some idea of the complexity of what you are trying to do.

I think many people here would agree that some patients with dementia have such erratic behaviour that you would find it difficult to establish a base line for some of them. For some there is no such thing as a regular pattern of behaviour and for others things can change gradually over a long period of time.

OH doesn't do much that a camera would observe as being different. His dementia is reflected through what he says and how he says it. If I asked him to do something different, he would be confused but then there's nothing to compare it with anyway.

 

[Jorjix]

I agree, this project needs to take into account different types of patients and I highly appreciate your inputs. The idea is to create something that actually helps carers and has a valuable impact.

The way I can see this system work with what CardiffLady described is that it would have to be taught by the carer that 'turn the place upside down, bed, mattress, fridge, freezer, empty cupboards, take up carpets, rugs, puts everything in the hallway' means 'panic' for a patient. Then it would be able to recognise it, if it occurs again and it would flag this sequence of events the way it was taught.

For what Lawson58 described, it might not be able to detect something in the movement, but it could pick anomalies in the speech (provided that the camera has a microphone), or maybe in the face expression. Although complex, this technology already exists and it's somewhat similar to how Google Translate is working.


Regardless of the difficulty of the attempt, lets imagine it is possible to implement a system that monitors a person in their home, over time, in the agreed spaces (living room, kitchen, halls, maybe bedrooms?, not bathroom). Lets say that the video is anonymised (blurred or contains only the silhouette of the person) and that the analysis is tailored to the monitored patient, with the goal of detecting vulnerability and falls risks, based on behavioural patterns.

If it existed, would such a system be of any use for the carer?


Regardless of what I described earlier, what kind of product would actually be valuable for you, according to your unique experiences? What would make caring for your loved ones a bit easier if solved by technology?

 

[Beate]

I don't think I would ever want to use an intrusive CCTV system, particularly when the person with dementia can't consent to it.

But seeing as we're aiming high here, have you ever seen the film Robot & Frank? I want the robot butler. He was great, and almost human.
http://www.imdb.com/title/tt1990314/?ref_=nv_sr_1

 

[Jorjix]

I don't think I would ever want to use an intrusive CCTV system, particularly when the person with dementia can't consent to it.

But seeing as we're aiming high here, have you ever seen the film Robot & Frank? I want the robot butler. He was great, and almost human.

Wouldn't the robot have two cameras instead of eyes and always be filming and listening, though?

 

[Lawson58]

I think your idea has the potential for some application but I just wonder if that might be a little limited considering the complexity of your project.

As you can see, there are some people who would be very reluctant to use it and then there are those patients who because they have 24/7 care either by family or in a nursing home wouldn't require it.

One issue I have is that for it to be really effective is that the system surely would need to be through the the whole house and that certainly raises the problem of personal privacy and dementia patients still have that right.

There is a lot of technology out there that has been adapted in a variety of ways to assist people with dementia but none of it has the analytic aspects of what you have envisioned. I would love to see you come up with something but perhaps you need to narrow the focus of your project initially but don't let that stop you from having a long term ambition.

Good luck.

 

[Jorjix]

Sure, I agree it would be quite difficult to build a product that's useful to every patient.
But my bigger concern is for it not to be valuable if it's possible in terms of technology.

It seems to me that the value that such monitoring might bring is lower than the frustration for giving away part of the privacy. Which is a perfectly fair point.

I'm still wondering how this perception might change in the future.
For instance, the new Amazon Echo has a camera integrated, so it might become a default in each house; the same way everybody has a camera nowadays in their pockets: theguardian.com/technology/2017/apr/26/amazon-echo-look-webcam-choose-fashion-outfits-alexa-smart-selfie-camera

 

[Kevinl]

OK, I've read the thread a few times and thought I'd stick my bit in as I think it's a great idea...but...
The very nature of AZ is that it's unpredictable, there are maybe ways you can monitor behaviour for the person that stands out like in an airport. The person that separates themselves from the "herd", looks round and particularly looks behind them a lot as they go to immigration or custom checks are the ones to watch, body language can give a lot away and in the hands of experts like you, probably more than people realise.
That said AZ is a whole different ballgame, in the home I never know if someone who picks up a cup of tea is going to drink it or throw it at me, I can walk past someone who may kiss me one day and kick me the next.
The "suspicious behaviour" does exist (in my experience) with people in the early stages, my wife used to move things, sometimes it was just her tidying things up other it was a sign that she'd start by just moving a few ornaments then stop, other times she'd end up throwing them at me.
I've sadly had the time on my hands to speak to a lot of people some of whom have decades of experience of dealing with people with mental health issues and the only things I've taken away are; anything that moves quickly is probably a danger, never trust anyone, only ever sit with your back against the wall, never focus on any one thing focus on everything and always expect the unexpected.
AZ is by its very nature irrational and unpredictable which is where the problem lies when you're trying to work out a predictive programme.
It's a bit of an oxymoron but people can be unpredictable in a predictable way like the smuggler at the airport, people with AZ are unpredictable...end of.
My advise would be to see if a local EMI home would let you spend a day shadowing a member of staff, talk to them then see what you think is viable.
Second issue (if you're not bored by now) is what happens if you detect a sign, do you text me 20 miles away in a meeting saying "your mum is looking a bit unstable", does an alert go to a care company who if they have someone available will want £25 call out charge and who pays for the false alarms? Would a care home pay for it and have a member of staff on constant alert running from signal to signal or would it be cheaper to pay someone minimum wage just to be sat there?
There are devise out there now that can monitor for falls but they're rubbish, lift your arm up then drop it quickly and it can registers a fall due to the sudden increase in G force or whatever.
I hate sounding negative but if one trained/experienced human being in a room can't predict what someone with AZ will do then personalised programming of how an irrational person behaves, individually, sounds like a hell of a project.
K

 

[NancySpain]

My parents both have dementia and live in their own house. Im keen on tech and have thought hard about it. My dad has conditions that lead to him falling with serious consequences.


Sent from my iPad using Talking Point

 

[Shedrech]

hello Jorjix
Kevinl has written eloquently on the unpredictability of those with dementia, and I do agree
that said, it is hard to diagnose dementia (especially given the various types and those much less often recognised early on)
I'm wondering whether your ideas might be able to help in the diagnostic process
I think I read that there is some kind of system to look at the movement of those with Parkinson's, which picks up certain early key changes
might your work build up a body of evidence on movement and behaviour to help diagnosis, in some way ...
just a thought

 

[carpe diem]

Hi. I've used a lot of great tech. I think your idea is good in theory but who is going to respond to the alerts. I live 4 hours away from my mum so I can't have any false alarms.
A system which detects something has already happened is of little use. Preventing falls or systems which don't allow the cooker to be left on or the sink to overflow are all available.
There's little point shutting the stable door after the horse has bolted.
I don't like to be negative but prevention is better than just alerts.

 

[Jorjix]

I've sadly had the time on my hands to speak to a lot of people some of whom have decades of experience of dealing with people with mental health issues and the only things I've taken away are; anything that moves quickly is probably a danger, never trust anyone, only ever sit with your back against the wall, never focus on any one thing focus on everything and always expect the unexpected.

Kevinl, thanks a lot, I can't agree more with your comment. As you mentioned, if a trained human can't predict what someone with AZ will do, it's very possible a smart system won't either.

While I know that some things can't be foreseen, I'm curious to look into things that can be picked up by a human and are predictable. My assumption is that if you had somebody constantly watching an AZ patient, they could understand changes in behaviour and prevent dangerous situations. For instance, I'm hoping that a carer that silently watched a patient 24/7 would be able to understand what the patient feels on a day to day basis, pick up changes that can be due to an infection, low mood, anxiety and act accordingly.

I know it sounds SF, but I'm exploring the idea of a personalised preventative healthcare. The current health system is treating falls, fractured hips, burns, rather than preventing them. An old person might be totally independent, but more vulnerable one day; they could fall, break their hip, end up staying 2 weeks in a hospital and lose their mobility, confidence and independence. I'm wondering to what extend it is possible to predict situations that might result in old people having to move out from their homes.

Second issue (if you're not bored by now) is what happens if you detect a sign, do you text me 20 miles away in a meeting saying "your mum is looking a bit unstable", does an alert go to a care company who if they have someone available will want £25 call out charge and who pays for the false alarms? Would a care home pay for it and have a member of staff on constant alert running from signal to signal or would it be cheaper to pay someone minimum wage just to be sat there?

I think it's a fair point to say that alarms might not be of any use and they'd probably create more anxiety for the family.

The point would be to say something like "90% chances for your mum to have an infection base on her walking patterns", or "your mum seems to have lost a lot of weight in the last month; you might want to investigate that" or "your mum is walking/exercising/eating more lately"...
People are not particularly good to ask for help and we sometimes wait for the last moment to do it, but that could be quite dangerous for old people. Also, we are biased and weight last week's experiences more than experiences that happened 3 months ago, so it might be hard for a carer to spot how fast things evolve with a patient.

But again, I'm not sure how helpful carers think this type of information might be.

 

[Jorjix]

hello Jorjix
Kevinl has written eloquently on the unpredictability of those with dementia, and I do agree
that said, it is hard to diagnose dementia (especially given the various types and those much less often recognised early on)
I'm wondering whether your ideas might be able to help in the diagnostic process
I think I read that there is some kind of system to look at the movement of those with Parkinson's, which picks up certain early key changes
might your work build up a body of evidence on movement and behaviour to help diagnosis, in some way ...
just a thought

Shedrech, thanks for your suggestions.
I thought about trying to early diagnose dementia. I think it's a good idea that might actually be a bit easier to implement, however, I started thinking of it's value and impact for patients & their families.
Would this be useful? Would it be interesting to, say, find out 2 years earlier than it's currently possible that one has dementia?

 

[Jorjix]

Hi. I've used a lot of great tech. I think your idea is good in theory but who is going to respond to the alerts. I live 4 hours away from my mum so I can't have any false alarms.
A system which detects something has already happened is of little use. Preventing falls or systems which don't allow the cooker to be left on or the sink to overflow are all available.
There's little point shutting the stable door after the horse has bolted.
I don't like to be negative but prevention is better than just alerts.

I totally agree and I'm currently imagining crazy ideas for closing the loop and preventing falls. I can't believe that, with so much technology going on (we're envisaging having holidays on the moon!) and falls being such a huge problem, nobody found a way to solve it.

To answer your question, I don't think the solution is to send an alarm saying "your mum is going to fall in the next 5 mins", but to try to understand changes in behaviour and assess risk on a daily/weekly basis.

 

[philamillan]

That system has potential but still requires a person at the end of the camera to interpret the behaviours.

There is power if one person could easily monitor 30 individuals and therefore appropriately interact with the family as well.

Privacy is a big issue but it seems impossible to have both safety and privacy unless we incorporate the technology of infrared cameras.

I have always thought that such systems could be outsourced for monitoring to other parts of the world in order reduce risk of abuse.

 

[Jorjix]

That system has potential but still requires a person at the end of the camera to interpret the behaviours.

There is power if one person could easily monitor 30 individuals and therefore appropriately interact with the family as well.

Privacy is a big issue but it seems impossible to have both safety and privacy unless we incorporate the technology of infrared cameras.

I have always thought that such systems could be outsourced for monitoring to other parts of the world in order reduce risk of abuse.

Ideally, the system wouldn't require a person looking at the videos, if it made accurate recommendations; but I agree, we're not there yet.

Personally, I think that the idea of infrared cameras is very clever: it's less invasive and it still offers the required information for making the analysis.
Another idea I'm investigating is around extracting the silhouette of the patient and throwing away the real video, together with integrating motion/radio sensors in the rooms where people are less likely to install cameras (such as the bathroom and the bedroom).

I really like the idea of outsourcing monitoring in order to filter false alarms!

 

[houdini]

houdini

Hi, I think where a monitoring system could be useful is in the situation our family is in where a relative has just had a diagnosis of AZ and is currently living independently with daily visits from friends and family. What is hard for us to judge is the point where more support is needed - overnight stays or 24 hour care. A system of monitoring could be useful - deciding what is monitored is the challenge. While we're on tech I would love to be able to monitor my relatives fluid intake (and urine output) - fairly convinced that they never drink water and so are often mildly dehydrated thus risking kidney and bladder problems.

I think monitoring should stop where we would all stop if actually present - this could be different places for different people but for most would be OK to know someone had got into bed or out of bed, or gone into bathroom, but not to go any further than that.