Post your face when it turned out most neuroimaging research results from the past two decades belong in the fucking trash.
http://www.pnas.org/content/early/2016/06/27/1602413113.full
>Functional MRI (fMRI) is 25 years old, yet surprisingly its most common statistical methods have not been validated using real data
>In theory, we should find 5% false positives (for a significance threshold of 5%), but instead we found that the most common software packages for fMRI analysis (SPM, FSL, AFNI) can result in false-positive rates of up to 70%
>a 15-year-old bug was found in 3dClustSim while testing the three software packages (the bug was fixed by the AFNI group as of May 2015, during preparation of this manuscript)
>It is not feasible to redo 40,000 fMRI studies, and lamentable archiving and data-sharing practices mean most could not be reanalyzed either
>>55422538
It's useless tech anyways. The brain is just an antennae to the soul so studying it is pointless.
>>55422538
Jesus Christ.
that's messed up
>>55422566
I like this type of humour.
the human brain can't even comprehend 0.5 frames per second so still images are pointless
>>55422538
those machines cost millions, what fucking monkey coded this shit
>>55422624
I've worked in the medical software industry for a short period of time, it's the most corrupt software development business there is. The company I worked for had 8 managers and one software dev. It was a "non-profit" and the managers each had 100k+ bonus salaries. I did dealings with several UK companies while there and they were even worse. This was months after a man named Michael Swann stole 17 million from the local hospital.
>>55422538
tip top kek
>>55422705
>>55422538
>was fixed by the developer in 2015
I bet they don't even know this bug exists in other shitty countries
>>55422538
I knew it, humans do only use 10% of their brain capacity
>>55422747
more like 2%
>>55422538
it's only one of the scan types tho
>>55422740
>>55422705
>>55422624
The bug isn't even the main problem if you check the article, just one of several major contributing factors to why the results are useless.
The article's conclusion is basically that people in the field need to sit down and figure out how to do this properly before pumping out more useless results.
>>55422775
It's the most widely used one by a wide margin.
>>55422797
is this your article? be honest
>>55422624
probably an indian
people think hiring indians will save money
can you explain me like I am 5?
>>55422817
The doctors used to to push you into a tunnel for fun and they found later that it wasn't as much fun as they were thinking they had so they wrote this article saying that instead of not having fun 5% of the time they didn't have it almost 70% of the time
>>55422538
That's what happens when you rely on C.
>>55422817
>a doctor with shitty glasses checks you out and says if you want to live he needs to cut off part of your brain
>you spent shitton of money on expensive treatment, rehabilitation, etc etc but at least you're alive you're happy
>15 years later someone checks the guy who diagnosed you and turns out he's blind and you weren't sick in the first place
>>55422851
That actually happened https://en.m.wikipedia.org/wiki/Walter_Jackson_Freeman_II
get fucked tumorfriends
>>55422805
No.
But I've written a similar one in a different field at one point. It was a smaller and less fundamental problem but it still revealed that quite a few people wasted lots of time and funds in follow-up studies based on borked results.
Science in general is riddled with crap like this tbqh.
>>55422817
Most scientists are bad at statistics, scientific software quality is terrible, the combination of the two factors rendered most of the results of the most popular neuroimaging technique useless. The fact the field is also bad on sharing data and archiving their approaches means most of the stuff isn't replicable/fixable with improved techniques.
>it's a miracle!
>no it's a bug
>>55422878
>Known for
>Popularizing the lobotomy
>Invention of the "ice pick" lobotomy
>>55422878
>m
kys
>Freeman was also known for being a bit of an oddball and he complemented his theatrical approach to demonstrating surgery by sporting a cane, goatee, and a wide-brimmed hat.
>>55422957
Triggered :^)
Sent from my iPhone
>itt /g/ thinks they know shit about neuroscience
fucking epic
>>55422991
>neuroscientist
>doesn't even get basic stats right
top kek m8
>>55422991
we know shit about horrid code
https://afni.nimh.nih.gov/pub/dist/src/3dClustSim.c(...)
if( do_athr_sum ){
int iathr ; FILE *fp ; char fname[256] ;
thresh_summer_athr(athr_sum_bot,athr_sum_top) ;
sprintf(fname,"%s.sumup.1D",(prefix!=NULL)?prefix:"ClustSim") ;
fp = fopen(fname,"w") ;
fprintf(fp,"# %s\n",commandline) ;
fprintf(fp,"# "
"1dplot -one -box -xaxis 0:0.1:10:5 -yaxis 0:0.32:16:2"
" -xlabel 'nominal \\alpha at fixed p'"
" -ylabel 'integrated FAR over p\\in[%g,%g]' "
" -ynames 1s:NN1 1s:NN2 1s:NN3 2s:NN1 2s:NN2 2s:NN3 bs:NN1 bs:NN2 bs:NN3"
" -x %s'[0]' -plabel '\\noesc %s' %s'[1..$]'\n" ,
pthr[athr_sum_top] , pthr[athr_sum_bot] , fname,fname,fname ) ;
fprintf(fp,"# alpha ") ;
fprintf(fp," 1s:NN1 1s:NN2 1s:NN3") ;
if( do_athr_sum > 1 ){
fprintf(fp," 2s:NN1 2s:NN2 2s:NN3") ;
fprintf(fp," bs:NN1 bs:NN2 bs:NN3") ;
}
fprintf(fp,"\n") ;
for( iathr=0 ; iathr < nathr ; iathr++ ){
fprintf(fp," %.4f ",athr[iathr]) ;
fprintf(fp," %.4f %.4f %.4f",
rfa_1sid_NN1[iathr],
rfa_1sid_NN2[iathr], rfa_1sid_NN3[iathr]) ;
if( do_athr_sum > 1 ){
fprintf(fp," %.4f %.4f %.4f",
rfa_2sid_NN1[iathr],
rfa_2sid_NN2[iathr], rfa_2sid_NN3[iathr]) ;
fprintf(fp," %.4f %.4f %.4f",
rfa_bsid_NN1[iathr],
rfa_bsid_NN2[iathr], rfa_bsid_NN3[iathr]) ;
}
fprintf(fp,"\n") ;
}
fclose(fp) ;
}
#ifdef USE_SHAVE
destroy_shave() ;
#endif
if( verb )
INFO_message("Clock time now = %.1f s",COX_clock_time()) ;
} /* end of outputizationing */
/*------- a minor aid for the pitiful helpless user [e.g., me] -------*/
(...)
>>55423078/*-------- run away screaming into the night ----- AAUUGGGHHH!!! --------*/
>>55423078
Looking through some of the files in that directory, is this what large-scale C programs look like? It's beyond description, how the fuck can anyone do anything with this?
>>55423236
What do you mean?
>>55423332
why god
>>55423112
Holy shit my sides have died
>>55423332
is this what happens when I get my brain scanned?
>>55423236
>is this what large-scale C programs look like?
No, it's literally just garbage code. There are some nice C codebases out there. Redis is a common example of good readable C.
>>55423236
It is the sort of code that C encourages to write.
>>55422585
What humour? It's true. The brain is nothing but a quantum entanglement component to our actual soul. Why do you think quantum entanglement exists otherwise?
>>55422878
Following his development of the icepick lobotomy, Freeman began traveling across the country visiting mental institutions in his personal van, which he called the "lobotomobile."
>lobotomobile
>>55423700
domo arigato mr loboto
>>55422566
But studying antennas isn't pointless.
>>55422878
>>55423700
>lobotomobile
I raughed
>>55423696
It stops being funny when you actually rope in things like that.
Quantum entanglement can't transmit information faster than light so if your theory is true it means that the soul is kept inside our heads.
>>55422812
15yrs ago tho.
Back then indians were like super intelligent robots and were worth hiring, now it's 90% trash.
>>55423792
>Quantum entanglement can't transmit information faster than light
SHALOM!!!
>>55423709
new meme
Post your face when the government can read your mind using neuroimaging but they pretend it's a completely useless technology.
Makes you relly think, if your wife was diagnosed with Multiple Sclerosis years ago and you still had the feeling that doctors just do wild guesses when looking at the images.
>>55422538
This is actually related to one of the research projects i'm working on.
Making a magnetic ion sensor to measure brain activity in MRI (NMR on people) instruments.
I can't tell you much more or I'll be shot.
>>55423792
>It stops being funny
only it never was.
>>55422566
>doesn't know about antenna engineering
>>55424045
Then why am I laughing?
>>55423332
>C programmers
>>55422991
Neuroscientist here. There isn't much to know tbqvhwy.senpai
>>55425427
What's your opinion on the OP article?
>>55422538
>>Functional MRI (fMRI) is 25 years old, yet surprisingly its most common statistical methods have not been validated using real data
Daily reminder that most things called functional usually aren't.
>>55425457
Functional programming?
>>55425448
There's basically three ways that you can get false positives in neuro data
1. Not correcting for multiple comparisons. Basically each test has a 5% chance of returning a false positive. If you run the test multiple times, that error rate inflates. It's like playing Russian roulette. The more you play the more chance you have of being shot. Some people don't correct for this and just re-run until they get the coveted p > .05. Google Atlantic salmon fmri for a laugh. This is not knowing what to do.
2. Errors in the implementation of tests. This is not knowing how to do something, or having the software mess up as in the article in the OP. Basically no one audits statistical tests in academic journals or in labs. No journal that I know of asks for your script and you barely have to describe how tests were conducted. Also almost no one knows how to code and everyone uses really shitty GUI packages like SPSS
3. Lying. This happens too.
Long story short, at least half of published neuro papers are either lying or at least misleading. Blame the HUGE lack of statistical training, the complete lack of understanding of programming, and the publish or perish incentive structure
>>55425587
>Basically each test has a 5% chance of returning a false positive
doesn't the article say something like the assumed 5% actually being "up to 70%"?
>>55425615
Here's how this part works - neuroscience (and psychology) run a test that is a comparison. If the things being compared are different, then we assume some conclusion. In fmri we're looking at blood oxygenization, which we assume is a measure of blood flow and therefore activity (oxygen is fuel so if a part if the brain is using more oxygen we think it's working). But brains are constantly working, so we have to take a snapshot of a specific area and compare it to other areas (and other brains) to see differences. The statistical test returns some values, including an estimate of how likely the numbers are to be essentially the same due to chance. If this number is below 5%, we assume they are sufficiently different and report it as "true". There's a difference so my treatment worked or this area does this or whatever my hypothesis was
>>55425737
The article is saying the implementation of these tests in the software is actually returning seemingly significant results at much higher error rate tolerance levels. In essence, we should be able to day we're 95% sure this is right but the software is saying we're something like 30% sure this is right. Then the authors go write papers like it's 100% certain and they get cited without being carefully read or properly reviewed in the first place
>>55425864
Details of the implementations is one thing, the more important problem described by the article is that the underlying assumptions for the statistical methods employed simply aren't true for real data.
>>55423236
No, it's just shit. I've worked on large C projects, they don't become like that without concentrated incompetence.
>>55422538
unsurprised, similar results have been found in a number of other research fields that haven't adopted six-sigma or other strong statistical safeguards against publication bias and false positives
>>55426463
>the more important problem described by the article is that the underlying assumptions for the statistical methods employed simply aren't true for real data Agreed. I can tell you in my program there are two required statistics courses, and the undergrad neuro program has literally 0
>>55423823
Do you have any argument besides calling your opponent a Jew?
So how many /pol/ arguments does this invalidate?
>>55427008
Everyone, prepare for habbening
>>55422747
>only use 10% of their brain capacity
You use all of your brain, just not all at once because it would cause a seizure.
>>55422538
Does this mean I don't actually have a brain tumor like they told me?
Is medicine the worst industry in existence?
>>55427159
I think we've come a long way, see >>55422878
>>55423792
That's literally what quantum entanglement implies. Ftl transfer of information.
>>55427144
No that's a structural MRI, not a functional MRI.
>>55427159
no, probably pharma
>>55427218
You should read a book
Or at least a wikipedia article, jesus.
>>55424222
Because you're insane
>>55427008
Well it invalidates most of the work of Sam Harris since he was using fMRI to invesigate how the brains of theists work differently from those of atheists.
Neuroscientists are generally lefties, /pol/ will cheer for this.
>>55422538
>lamentable archiving and data-sharing practices
I've worked in hospitals. You'd cry if you knew how much stuff is still stored on paper, and there's only one paper copy of it and there are rooms just full of paper and files and nobody is even sure who 50% of the files belong to.
>>55422890
>Science in general is riddled with crap like this tbqh.
That's what I figured...
>muh citations
>muh pyramid scheme
>>55425566
Good example.
The worst thing about this it's going to take decades to spread awarness that basically all we thought we knew about the brain through this is complete bullshit. And trying to dispel whatever myths may have been cause by it.
I actually work in a hospital. I can't honestly think of a patient that just gets an mri and goes to the or. That's not how shit works.
There's a whole battery of shit done first. What they're complaining about, if they're symptomatic, vitals, ct scan, ecg/ekg, etc. Mri might have false positives, but it's taken into consideration with everything else.
>>55427348
Most computers in hospitals still run on Windows XP without updates
>>55422624
>>55422705
Medical software is even worse than web dev and game dev code monkeying. A coworker of mine has worked in medical software in the past, and his horror stories are brutal.
>>55427686
This isn't primarily about clinical applications, the whole body of research based on fMRI just got its foundation demolished.
>>55422878
>This new procedure became known as the "icepick" lobotomy and was performed by inserting a metal pick into the corner of each eye-socket, hammering it through the thin bone there with a mallet, and moving it back and forth, severing the connections to the prefrontal cortex in the frontal lobes of the brain.[9]
and I must scream
>>55422878
>mfw reading about Howard Dully
I want to punch the shit out of his bitch stepmother.
>>55427309
I've seen /pol/ theists use it to argue atheists are brain-damaged, most arguments about neurology/psychology relating to ideological preferences can be twisted either way.
>>55422878
To think, the mental institutes were crowded because parents would send perfectly normal semi-rebelious teens there thinking they were supposed to come out of the womb perfect angles and remain that way till they were 18. Like the last/current fad was ADD/ADHD meds but at least that made people insane at studying so they could have an easy time through college.
>>55427443
Citations are themselves useful to prevent basing research or premises on "everybody knows it" or other dubious nonsense, but citation systems are basically academic autofellatio.
Isn't it bad to blast yuor brain with radiation?
