“...the equivalent of 1,700 years of research data on Alzheimer’s.”
April 29, 2019 9:34 AM   Subscribe

A Video Game Developed To Detect Alzheimer’s Disease Seems To Be Working [Kotaku] “Sea Hero Quest is a video game developed in partnership with Germany’s Deutsche Telekom, game studio Glitchers and several European universities and it is designed to identify individuals who might have early and mild symptoms of dementia that medical tests aren’t able to detect. [...] In Sea Hero Quest, which is a VR game, players have to navigate and control a virtual boat. They are given a map and shown checkpoints, then the map is taken away and players must navigate to these checkpoints in the game world without the map. According to researchers, every two minutes spent playing the game is equal to five hours of lab-based research.” [YouTube][Game Trailer]

• A game that explores the effects of Alzheimer's Disease [Engadget][YouTube][Game Trailer]
“You're standing in a living room. It might be your living room. Something's not quite right -- for every object you recognize, there's another you don't. Normally, if you get this feeling while playing a video game, you're in the middle of a horror, awaiting the inevitable jump scare. In Forget-Me-Knot, however, you aren't evading an enemy, but instead trying to piece together memories of a life that, thanks to Alzheimer's Disease, you barely remember. Developed by Alexander Tarvet, a student studying Game Design & Production Management at Abertay University in Dundee, Scotland, Forget-Me-Knot is designed to raise awareness for those suffering with dementia. [...] "The player is in exactly the same position as the person with Alzheimer's -- both have to explore the room and try and piece together an understanding of photos and letters through clues left on shelves and in drawers."”
• Before I Forget: the video game that tackles dementia [The Guardian][YouTube][Game Trailer]
“When you’re playing Before I Forget, wandering around a house inspecting everyday objects, trying to trigger a memory, it feels like a lot of narrative-driven video games: you’re attempting to piece together a story using clues left behind. But this game puts you in the shoes of Sunita, a woman suffering from early-onset dementia, which infuses what you’re doing with sadness and significance. The house is delineated in monochrome, colour seeping back as she gradually reconnects with her past self. Examining a photograph provides a clue to her identity; a familiar piece of music might recall an important person in her life. Other symptoms are conveyed in more disconcerting ways. Sunita can become lost in her own home in a nightmarish loop, opening doors that all lead to the same dark broom cupboard. Before I Forget was exhibited as part of The Leftfield Collection’s experimental indie-game lineup at Rezzed games convention in London earlier this year. Developer 3-Fold Games hopes to provide a sensitive and emotive portrayal of dementia.”
posted by Fizz (21 comments total) 34 users marked this as a favorite
 
Thanks. Gonna download this right now. (Hope that dreadful music doesn't play when you're playing it.)
posted by dobbs at 10:26 AM on April 29, 2019


From the Kotaku article, a quote from the researchers:

"We found that people with a high genetic risk, the APOE4 carriers, performed worse on spatial navigation tasks."
While I'm not really sure what else I could ask them to do, other than longitudinally follow these folks and report back in twenty years... This doesn't seem to be like a good test. This is one gene that's been linked to Alzheimer's, but to my knowledge, not mechanistically, and inferring that the poor performance of these carriers is evidence that the game can detect early Alzheimer's seems really really dubious.

Detecting a significant difference in performance in a population doesn't seem to be to be the same thing as creating a diagnostic instrument. And I'm not even going to comment on the lab hours equivalency statement, it's impossible to clarify what that means.

This is a really neat post, though. I'm really interested in the simulation games, they seem heart breaking and good.


And I will try to connect
All the pieces you left
I will carry it on
And let you forget
And I'll remember the years
When your mind was clear
How the laughter and life
Filled up this silent house
Silent house

posted by Made of Star Stuff at 10:35 AM on April 29, 2019 [4 favorites]


For years, my friends, family, and coworkers have had to listen to me rave about the scientific and teaching applications of VR. Don't get me wrong, immersive VR games are tons of fun and that's MY primary reason to have a VR headset but it's really all leading the way to having tangible things that further us.

I may be biased as someone who learns much better visually, but this type of stuff is exactly what I've been going on about.

Even if this doesn't end up being a solid method, this kind of thing is the future of learning and discovery. I' so excited.
posted by Twain Device at 10:47 AM on April 29, 2019 [2 favorites]


I'm sure I will be downloading the Glitcher's game and playing it but frankly I could do without all of the humbug. The 2 minutes equals 5 hours line is really really off putting.
posted by Pembquist at 11:04 AM on April 29, 2019 [1 favorite]


I've been on a conference call or two with one of the researchers (Hornberger) behind this; it's fascinating work, and anything that might help with creating some tools and methods for early detection is great.

This doesn't seem to be like a good test. This is one gene that's been linked to Alzheimer's, but to my knowledge, not mechanistically, and inferring that the poor performance of these carriers is evidence that the game can detect early Alzheimer's seems really really dubious.

As is often the case in science reporting (and I find it particularly bad in the Alzheimer's/dementia world, where the media proclaims every new development as some level of "silver bullet", but that's probably because of my closeness to the work), the paper's claims look to be more modest than what the media grabs onto. The study is suggesting that the game can help to sort people into higher-risk vs. lower-risk groups, allowing for better follow-up in terms of further diagnostic workups and so forth, not that it is the best diagnostic tool of all time. They used a sample with the APOE4 gene as a way of identifying a higher-risk group in order to establish that the game can detect differences in such a group versus a "normal" population.

I'm off to spend some time with the full paper and see what it says.
posted by nubs at 11:05 AM on April 29, 2019 [10 favorites]


Well I guess I won't be playing Sea Hero Quest as it doesn't seem to exist in the IOS world, maybe version 1 is over and it is just Samsung VR now?
posted by Pembquist at 11:28 AM on April 29, 2019


I assumed that the whole "2 minutes = 5 hours in the lab" thing referred to the quantity of data collected -- i.e. that gameplay generates data at approximately 150 times the pace of traditional psychometric methods. Which seems plausible, but I have no idea how accurate/relevant that is. (The paper itself doesn't seem to make this claim, so maybe it's something that AAAS or a university PR person came up with?)

Of course the problem with claiming that the game data is equivalent to 1700 years of lab data is that there's no longitudinal component -- the game can generate lots of short-term data quickly, which hopefully can turn into lots of long-term data if people keep playing the game regularly for the next 20 years, but actually getting that longitudinal data is going to take exactly as long as it would with traditional "small data" methods.

Big Data is cool and all, but it still hasn't disrupted the space-time continuum.
posted by shenderson at 11:39 AM on April 29, 2019 [4 favorites]


Pembquist, it looks like they must have removed the iOS version -- I played a few hours of it last year on my iPhone.
posted by matrixclown at 12:17 PM on April 29, 2019


I look forward to performing poorly on a video game, and then, as a result, being declared incompetent by the courts.
posted by amtho at 12:45 PM on April 29, 2019 [5 favorites]


The 2 minutes equals 5 hours line is really really off putting.
It reminds me of Duolingo's in-app PR nonsense.
posted by aspersioncast at 1:07 PM on April 29, 2019


Big Data is cool and all, but it still hasn't disrupted the space-time continuum.

Not for want of trying.
posted by Celsius1414 at 2:39 PM on April 29, 2019 [1 favorite]


OK, having read through the paper once here are some notes. First, some generalities:

-We know, in general, that the physiological changes in the brain that come with AD start decades before there are symptoms that would lead to a diagnosis with current testing. With AD, memory loss is generally considered to be one of the first (and hence best known) symptoms.
-One theory about why the medications we have for AD have limited effect is that we administer them too late; that their effectiveness might be greater if we could start giving them to people even 10-15 years before we can right now because diagnosis right now depends on the symptoms reaching a point where the symptoms are causing problems. This is why there is a lot of work being done on trying to improve the sensitivity of diagnostic tests and find methods of diagnosing much earlier, at what would be called the "pre-clinical stage" or pre-symptomatic stage, basically meaning that there might be deficits in performance, but they do not impact the person's ability to live their lives day-to-day in a meaningful way. This would allow for earlier intervention.

Now, from the paper. Anything that is me adding something that isn't in the paper is in italics:

-spatial navigation is emerging as a potential early marker for diagnosis, one that may appear ahead of memory problems.
-However, age and gender also affect spatial navigation - so any tools that try to make use of spatial navigation in the diagnosis of AD need to be able to take that into account.
-Further, the issue of identifying who is at high-risk in pre-clinical/pre-symptomatic studies is difficult; this paper uses the APOE4 gene as the best indicator of high-risk for sporadic AD (this is the non-familial/non-inherited form of AD, which involves either the APP, PSEN1 or PSEN2 genes). People with the APOE4 gene are noted to have errors during wayfinding.
-the study sets out to determine if the Sea Hero Quest app can replicate the wayfinding difficulties in APOE4 carriers; if it can handle sex, age, genetic differences on spatial cognition; and if AD-specific wayfinding problems can be detected in individuals, when compared to a benchmark comparison of data from app users.
-the study compares the performance of 31 APOE4 users versus 29 APOE3 users versus 27,108 benchmark players on the app; the demographics were the same for the two groups (I don't have access to their appendices so I'm taking their word for it), and they scored similarliy on neuropysch exams (the Addenbrooke's cognitive exam, Rey Complex Figure Task, and a recall task on Rey Complex Figure Task).
-APOE3 carriers travelled shorter distances during wayfinding than APOE4 carriers, to a statistically significant level, when controlled for age and sex. There was no significance found for time on the wayfinding task.
-APOE 4 carriers took different routes, tending towards the the border of the environment, while APOE3 carriers navigate along the centre of the virtual environment.
-no effects of sex were found on distance travelled, but men took less time than women. No significant effects were found when looking at the combination of geneotype and sex - in other words, this finding appears to have no bearing on the risk level for either sex.
-no significant effects were found for age on the scores in the app.
-they also compared the groups on a memory task and wayfinding ability, and found that the results of the wayfinding task predicted the presence of the APOE4 gene better than the memory task did.
-they conclude that their test showed the app is replicating previous studies that show APOE4 gene carriers have wayfinding differences; that sex differences are present in wayfinding time but that the effect of sex is negligible compared to the genetic factor of the APOE4 gene; and that people "at genetic risk" of AD with no memory deficits can be distinguished on wayfinding measures at an individual level.
-because of the fact that AD affects women more than men, noting the sex differences is important and more research is needed to see if sex differences in navigation ability are important, particularly in situations where the genetic component may be unknown.
-normative measures are needed to accurately assess spatial navigation problems, as this is not yet well-established as an area of assessment, so that a standardized test can be created. The app may provide unique benchmarking data, as it can provide information about sex, age, and cultural background, which will be important. Education data has also been collected, but has not shown any effect on navigation performance in the full database.
-the app includes some tutorial levels, so that video game familiarity/ability could be controlled for - data from these, and the first few levels were also excluded so that players were comfortable with the mechanics/controls.
-They conclude by noting that their data from the app may help in the creation of a future test around spatial navigation and that navigation may be a complimentary biomarker to the others that are used for future AD diagnosis and interventions.

So...some interesting stuff here (my immediate response to this is that the sample size is low, but that is very common in AD research for emerging/new possibilities). Certainly, the study is not what the media is claiming about this; the authors are working from the theory that spatial navigation is an important early diagnostic marker (not the only one), that an app like this can help detect problem in high-risk groups, and suggesting that there is a need for more study and especially data regarding what the baselines are for different populations so that this could be used alongside other diagnostic tools to improve identification and decision making about intervention. They also think that the app offers a way to help collect that baseline data to help with the development of a possible standardized test around this, that also is sensitive to differences in sex, age, culture, education, etc.

There's no mention of this "2 minutes equals 5 hours" weirdness in the paper that I saw, so I don't know where that is coming from.
posted by nubs at 3:04 PM on April 29, 2019 [13 favorites]


Sorry, just to add a quick note:

The APOE gene comes in some variants - the APOE4 gene is considered to increase risk of developing AD; APOE3 is considered neutral in terms of risk; and the APOE2 gene may reduce people's risk.

In sporadic AD (the vast majority of cases), there are a large mix of risk factors at play - gene variants, such as APOE, but also factors like age, gender, education levels, cardiovascular health, diet, sleep, mental health, diabetes, and a bunch of others. In short, AD (and other dementias) are complex and we are still learning a lot about them.

I can go on and on, but I will shut up now.
posted by nubs at 3:14 PM on April 29, 2019 [2 favorites]


In addition, there are some really promising new drugs that may halt brain mass loss! It's looking up.

OTOH, while discussing the new drugs and reliable indicators with Dr.es_de_bah (Ms. es_de_bah is a psychologist. Please forgive me as I layman paraphrase the convo). An interesting factor came up: There are two physical traits that obviously distinguish older brains from younger ones. Most obvious and relatively well understood is Alzheimer's. Amyloid plaques build up, neurons die. Lost connections.

The other potential problem* comes from the increasing strength and resilience of connections in old brains. One prevailing theory is that your brain makes connections by reinforcing neurological pathways as you re-associate certain ideas, memories, prejudices thoughout your life. This may be why higher-functioning Alzheimer's patients fall the hardest and fastest all at once. Strong, healthy pathways are resilient until they aren't, and it may be more difficult to create new pathways because of it. That is, older minds become more inflexible due to strong, well reinforced neuron connections/pathways. As our population lives to greater and greater ages, and if Alzheimer's can be made rare, we may become more and more familiar with an elderly memory and capability issue that has more to do with an ever-more inflexible mind living long into an ever-changing world.

*Again, this part is still pretty hypothetical, so I hear and understand. Folks who no better, please correct me if I'm talking shit.
posted by es_de_bah at 4:11 PM on April 29, 2019 [1 favorite]


Metafilter: it still hasn't disrupted the space-time continuum.
posted by Greg_Ace at 4:57 PM on April 29, 2019 [2 favorites]


I used to play Sea Hero Quest, and the game used to insist that it wasn't intended for diagnostic purposes.
The stated goal of the game was to compile a baseline of "average" navigation ability for comparison purposes.
posted by cheshyre at 6:53 PM on April 29, 2019


In the future, I will go into an office and maybe be invited to play a video game which will decide my agency & future. That's just the way it is.
posted by ovvl at 6:58 PM on April 29, 2019


>The study is suggesting that the game can help to sort people into higher-risk vs. lower-risk groups, allowing for better follow-up in terms of further diagnostic workups and so forth

I’m sorry, it’s true, that’s the aim, it’s a good aim. I just find this idea amusing. What diagnostics? In the words of one horrorshow of a doctor I had the misfortune of dealing with, “Oh anyone can diagnose it in ten minutes, you can find the test on the internet. Your parent? Yes, probably demented. This happens when people get old, not sure what you expect. Oh no, there’s nothing for it.” He was the absolute worst, but the other assessments, by (thankfully) a non-asshole, took under an hour. And in the end, the jerk was right, there’s no real help for any of it. (AFAIK, the drugs that are meant to delay progression only work for some. For maybe a year.) At least right now - please, *please*, correct me, please. (The #1 thing a neuropsych prof of mine said it’s thought can help in terms of *prevention* is exercise. Which felt slightly empowering but more inadequate.)

Sorry. I’m very glad researchers are working on interventions and diagnostic tools. I hope they make a difference.
posted by cotton dress sock at 7:17 PM on April 29, 2019


That's just the way it is.

It is if everybody keeps saying "That's just the way it is."
posted by amtho at 5:12 AM on April 30, 2019


What diagnostics? In the words of one horrorshow of a doctor I had the misfortune of dealing with, “Oh anyone can diagnose it in ten minutes, you can find the test on the internet. Your parent? Yes, probably demented. This happens when people get old, not sure what you expect. Oh no, there’s nothing for it.” He was the absolute worst, but the other assessments, by (thankfully) a non-asshole, took under an hour. And in the end, the jerk was right, there’s no real help for any of it. (AFAIK, the drugs that are meant to delay progression only work for some. For maybe a year.) At least right now - please, *please*, correct me, please. (The #1 thing a neuropsych prof of mine said it’s thought can help in terms of *prevention* is exercise. Which felt slightly empowering but more inadequate.)

You had an asshole jerk of a doctor, and I'm sorry. Without knowing more about your situation, here's what The diagnostic tests that he was referring to are likely things like the Montreal Cognitive Assessment (the MoCA); they do detect changes in cognitive ability, but shouldn't be used as the exclusive method for diagnosis of dementia because there can be other things at play that might be causing those issues. And one of the things with them is that they are good at detecting changes in cognitive ability after they become a problem with someone's day-to-day life; research like what is in the study here is aimed at seeing if we can detect the development of issues earlier than what we have currently, precisely because by the time issues are detected with the current tools there are no effective treatment options.

The drugs that currently exist are not great, by any means - for many people, the effects are modest, and they don't have long windows of effectiveness either. But it may be that those drugs are being applied too late in the disease process to make a difference; the physiological changes in the brain have been happening for decades by the time they are detected by things like the MoCA. So, if we could detect issues much earlier, those drugs may be more effective or there might be other strategies to try. And, yes, exercise is one of the best things we know about as far as a preventative/risk-reduction strategy at this point; regular physical activity is important for a lot of things for health, including our brain.

This study is aiming to see if there are techniques that can be used to detect higher risk levels earlier, which (if it is possible) would mean that they can consider what approaches they might take to lower their risk - like modifying their lifestyle with things like exercise and diet and other possibilities, just like people who have indications of elevated risk for other diseases.

Anyways, I don't know if that helps or not, and I'm sorry about what's happening with your parent, cotton dress sock.
posted by nubs at 7:48 AM on April 30, 2019 [1 favorite]


I think what irks about this game/data collection is not the idea of it but the marketing part, the part with the 2 minutes equals 5 hours the part where one of the collaborators is Saatchi & Saatchi. So it is a little bit of an effort to leave all of that aside but leaving all of that aside, one of the things that I have observed with regard to dementia/alzheimer's is that there is a distinct unwillingness to engage in the subject among many doctors. In general my experience bringing up concerns about my memory etc. with a primary care doctor engenders a dismissive or "reassuring" response the take away from which seems to be that there isn't anything you can do about it so why would you want to know and we cannot tell you anyway. This is not just my experience but everyone I know that has had similar concerns. This leads me to the idea that if instead of just getting a shrug from a doctor what would make more sense would be an offer to participate in a long term large data collection effort that would leverage the new tools and costless scalability of the internet and personal computing devices. The idea would be to have both a database of verifiable subjects and an armature on which you could implement studies like the one that this video game was part of instead of having to reinvent the wheel each time. The data generated generated by each additional study would feed back into the database drawing a more and more complete picture of an individual subject and generating more and more metrics for comparison between a growing and diversifying collection of people. The data would have to be both anonymous yet verifiable, available for research and not locked inside some startups IP vault waiting for the day when they could "make the world a better place," which is frankly the vibe Sea Hero gives off.

I can guarantee that you would get a very high rate of participation from anyone who has had a loved one stricken.
posted by Pembquist at 9:13 AM on April 30, 2019 [1 favorite]


« Older A lot of people in that building were pretty...   |   The Consultant Crisis Newer »


This thread has been archived and is closed to new comments