4 ms·

Autism Diagnosed with a Fifteen Minute Brain Scan

20 points by kkleiner 16y ago

Eliezer 16y ago
So I'm reading and thinking, "this information is useless, I can't get a likelihood ratio out of this"... and then someone actually provides the Bayesian posterior probability! Woohoo! Rationality prevails! Could this be the dawn of a new era?
- pjscott 16y ago
  The guy who calculated the likelihood ratio was writing in a blog post in response to a bunch of mainstream articles which failed to do so. And then he was quoted in SingularityHub, which I would hope would be a bit better-informed than typical science journalists. If this is the dawn of a new era, then the defining trait of this new era is that people outside the mainstream can make themselves heard, thanks to the internet. I'll happily change my mind about that if I see someone like the BBC mention posterior probabilities in an article about this.
MJR 16y ago
How annoying that they continue to refer to Autism as a disease. Autism is a developmental disorder.
- grannyg00se 16y ago
  Some would claim even further that it is neither a disease nor a disorder of any kind, but rather simply a differernt type of brain that should not be considered diseased, disordered, or broken in any way. Just different. Something to be accepted and learned from.
- AngryParsley 16y ago
  When talking about whether obesity or autism or whatnot is a disease, I think this post sheds some light on the topic: http://lesswrong.com/lw/2as/diseased_thinking_dissolving_questions_about/ http://lesswrong.com/lw/2as/diseased_thinking_dissolving_que...
StavrosK 16y ago
At a presentation of my university (UCL), one of the presenters told us that, amazingly, she was able to detect clinical depression from an MRI scan with very high accuracy. To be fair, she was not able to detect mild cases, but I don't think the doctors can really distinguish between people being depressed because of an external and a physiological cause... That was the most surprising thing I've heard in quite a while...
rtp 16y ago
Ah, this one. Say you have a sample group with the size of 10000 people (all randomly picked). The prevalence of autism is calculated to be around 1% of an typical population. This would mean that the group consists of about 100 people who are autistic. The accuracy of this test is 90%, and thus this would yield a positive result of about 90 people who really have autism, and about 1000 people who don't.

20 points by kkleiner 16y ago

So I'm reading and thinking, "this information is useless, I can't get a likelihood ratio out of this"... and then someone actually provides the Bayesian posterior probability! Woohoo! Rationality prevails! Could this be the dawn of a new era?

pjscott 16y ago
The guy who calculated the likelihood ratio was writing in a blog post in response to a bunch of mainstream articles which failed to do so. And then he was quoted in SingularityHub, which I would hope would be a bit better-informed than typical science journalists. If this is the dawn of a new era, then the defining trait of this new era is that people outside the mainstream can make themselves heard, thanks to the internet. I'll happily change my mind about that if I see someone like the BBC mention posterior probabilities in an article about this.

MJR 16y ago

How annoying that they continue to refer to Autism as a disease. Autism is a developmental disorder.

grannyg00se 16y ago
Some would claim even further that it is neither a disease nor a disorder of any kind, but rather simply a differernt type of brain that should not be considered diseased, disordered, or broken in any way. Just different. Something to be accepted and learned from.
AngryParsley 16y ago
When talking about whether obesity or autism or whatnot is a disease, I think this post sheds some light on the topic: http://lesswrong.com/lw/2as/diseased_thinking_dissolving_questions_about/ http://lesswrong.com/lw/2as/diseased_thinking_dissolving_que...

StavrosK 16y ago

At a presentation of my university (UCL), one of the presenters told us that, amazingly, she was able to detect clinical depression from an MRI scan with very high accuracy. To be fair, she was not able to detect mild cases, but I don't think the doctors can really distinguish between people being depressed because of an external and a physiological cause... That was the most surprising thing I've heard in quite a while...

rtp 16y ago

Ah, this one. Say you have a sample group with the size of 10000 people (all randomly picked). The prevalence of autism is calculated to be around 1% of an typical population. This would mean that the group consists of about 100 people who are autistic. The accuracy of this test is 90%, and thus this would yield a positive result of about 90 people who really have autism, and about 1000 people who don't.