News for hackers. Stuff that matters.
Re:What Killed My Sister? by Goopplesoft on Monday April 28, 2014 @08:59PM (#7664203)
> because we have no blood test or genetic test for it

I was reading about this: http://ge.tt/3nXdgiQ/v/0, its from Reddit where the article stated:

"However, when we examined the whole data set of 298 assessments, a cross-validated probabilistic neural network model was superior and could discriminate all cases from controls with near perfect accuracy at 98.3%."

Is this not the case?

(article here: http://www.scotsman.com/news/health/schizophrenia-diagnosed-...)

Re:What Killed My Sister? by pdpi on Tuesday April 29, 2014 @07:17AM (#7666193)
Depending on what they're measuring, 98.3% can be pretty bad. Given that there's about a 1% portion of the population with schizophrenia, if the test is right 98% of the time (irrespectively of everything else), then it'll misdiagnose two healthy people as having schizophrenia for every correct diagnosis.
Re:What Killed My Sister? by mercer on Tuesday April 29, 2014 @08:07AM (#7666354)
Yes, but wouldn't it be obvious (or obvious eventually) that these healthy people clearly don't exhibit other symptoms of schizophrenia? Or am I forgetting something?
Re:What Killed My Sister? by acdha on Tuesday April 29, 2014 @08:26AM (#7666437)
What about terrifying people with the thought they have schizophrenia but just haven't demonstrated visible symptoms yet? That concern comes up a lot with genetic testing because many things aren't certain but people panic and make significant life-altering decisions after a positive test result.

EDIT: … I should also have noted that humans are notoriously suggestible – with a mental illness it's really easy to imagine people both assuming every quirk is proof of a terrible disease or even starting to demonstrate symptoms psychosomatically after being told that this will happen eventually.

Re:What Killed My Sister? by mercer on Tuesday April 29, 2014 @08:31AM (#7666456)
Good point. I suppose it should be limited to people who are already showing other symptoms of schizophrenia...
Re:What Killed My Sister? by acdha on Tuesday April 29, 2014 @09:37AM (#7666847)
… or with family histories, etc. The real challenge, though, is probably just for doctors to be very open about false positive and negative rates and explain the issue directly. I think a lot of these issues would be easier if someone was frankly told “3 times more people receive positives on this test than actually have the disease“ instead of “it came back positive for [scary]”.
Re:What Killed My Sister? by VonGuard on Tuesday April 29, 2014 @12:25PM (#7668043)
Nevermind the classic problem that was addressed in One Flew Over the Cuckoo's Nest and others: how do you prove you're sane? If the machine says you're insane, what way is there to prove otherwise?
Re:What Killed My Sister? by andreasvc on Tuesday April 29, 2014 @10:57AM (#7667380)
You are assuming errors would be random, which is probably false. A false positive would probably be more likely for someone with similar symptoms, than for someone without any symptoms.
Re:What Killed My Sister? by caycep on Monday April 28, 2014 @09:07PM (#7664239)
Hm, that's interesting. A lot of parkinson's related disease (different than schizophrenia, but still involving dopamine a lot) have eye movement findings that are pretty recognizable.

Culturally, not sure if psychiatrists are in the habit of examining eye movements as closely as, say, a neurologist or neuro-opthalmologist, but it'd be interesting if this is replicable by other groups.

Re:What Killed My Sister? by stan_rogers on Tuesday April 29, 2014 @01:29AM (#7665145)
It's only another tool in the diagnostic toolkit. My presentation is, um, "non-classical" to say the least (intentional tremor being, perhaps, the weirdest part of it) and arriving at a diagnosis of Parkinson's disease pretty much involved ruling out everything else that might have been causing my symptoms (which is a rather long process, all things considered). I've met dozens of people who I could have spotted as Parkinson's sufferers two blocks away at night during a blackout, and a few who, like me, seem to have a bespoke version that "isn't the other things" and responds to treatment for Parkinsonism. Well, as well as anything does, at least - meds are always a pretty blunt weapon and can't exactly track the body's own chemistry, so there will always be good periods, and under/over-medicated periods as well. When it comes to neuropathy - or normal functioning, for that matter - we're not quite stumbling around in the dark, but there's still one hell of a long way to go before we can say that we understand much about the machine.
Re:What Killed My Sister? by sitkack on Monday April 28, 2014 @09:36PM (#7664366)
There was another test involving anomalous speech patterns, http://www.idt.mdh.se/kurser/ct3340/ht11/MINICONFERENCE/Fina...

it might not be the canonical work.

In the near future most laptops will have the ability to do realtime eye tracking so these kinds of tests could be administered broadly, probably even through a web page.

Re:What Killed My Sister? by caycep on Monday April 28, 2014 @10:06PM (#7664485)
how close are consumer eye tracking devices? It'd be interesting to see. My group works w/ the PD side of things, and doing a desktop-based eye test might be something we'd be interested in...
Re:What Killed My Sister? by sitkack on Monday April 28, 2014 @11:41PM (#7664808)
I have used an eye tracker in about 5 years but I think the cost back then was on the order of 4k+-1k.

http://www.tobii.com/en/eye-experience/ this is probably better than what we were using.

So I'd say all the pieces are in place to do this. The architectural design of the tobii and the leapmotion is almost identical. High frame rate cameras over USB3 illuminated with IR.

----

Throw in an http://emotiv.com/ EEG and one could do all sorts correlations.

Re:What Killed My Sister? by caycep on Tuesday April 29, 2014 @11:45AM (#7667742)
What I don't like about Emotiv is that external EEG is never that great - it only gets aggregate LFP's from signals from > 6 square cm of cortex at a time, and high frequency signals don't get through the bone and muscle. Still, data is data....and the sad part is neuro people have been looking at EEG for over a century and never really bothered to do real signals analysis on them...
Re:What Killed My Sister? by sillysaurus3 on Monday April 28, 2014 @09:34PM (#7664351)
I was reading about this: http://ge.tt/3nXdgiQ/v/0, its from Reddit

Would you mind linking to the Reddit thread that it was mentioned in? Just curious.

Re:What Killed My Sister? by Goopplesoft on Monday April 28, 2014 @11:04PM (#7664679)
Totally forgot about that and to refesh on the comments too. Here ya go: http://www.reddit.com/r/science/comments/12gxgs/schizophreni...