Human Connectome Project Updates | Hacker News Reader

Human Connectome Project Updates

spectrum.ieee.org

45 pointsneurotech128 comments

The article title is rather clickbaity. The only unsettling aspect is this:

> The HCP project has also moved brain scanning into the realm of the feature film Minority Report by showing that a person’s brain activity is as unique as a fingerprint and that it can be used to identify a person with 99 percent accuracy.

Once someone can scan your brain, they have many other ways to determine who you are. I was expecting something substantive about brain structure/function.

We can all envision the hackers coming, armed with an Fmri scanner at hand.

With a 22MW fusion pack ;)

And actually, for biometric data, 99% is fairly crappy.

After, a person's birthday identifies them with a greater than 364/365 accuracy.

If you give a color badge to each person in the world and you have 365 colors to choose from, how accurately would a specific color identify a person?

What is 1-(1/365)?

Should probably involve the population there somewhere?

Sort of.

If I have a population of 1, or <= 365 and uniform dist of birthdays, the probability of a date being a valid identifier is 100%.

If I have a population of 365 with randomly distributed birthdays, then it is going to be the probability of not sharing a birthday with any of them (very low)

Once you can guarantee at least 2 people with each birthday the identifier is simply zero percent effective.

Wait, what about leap years?

> radiologists to acquire high-resolution images up to 8 times as fast as they previously could using traditional MRI machines.

I think the author means 'radiographer' not radiologist. I also wonder about the claims as to how this will change clinical imaging. I get that this is a a amazing project, and it will have some far reaching effects but changes to clinical imaging will be limited imho. Multibanding existing sequences will make them faster, but sequences that benefit from the technique (bold and diffusion primarily) are not big clinical techniques. Diffusion is a routine tool but at <2 minutes it's hardly a problem to acquire. Maybe I'm missing something big like everyone getting a a functional scan as part of a routine clinical brains scan?

> Diffusion is a routine tool but at <2 minutes it's hardly a problem to acquire.

The kinds of high-angular coverage scans they are talking about here (60-120+ gradient directions) take 45+ minutes on current clinical scanners. These scans give improved white matter resolution -- both spatially and for complex sub-voxel microstructure. With acceleration and faster gradient ramps the connectome scanner can get that down to under 15 minutes. It's still a long time for routine clinical use (typical U.S. scan sessions are billed in ~40 minute increments), but more feasible than 45+.

We are running scans with 64 directions that are 8ish minutes, they look good. I'll have a look when I get into work, because there is obviously a large difference between their sequences and ours. Ours are not product sequences but they aren't anything too radical. Edit: to step further back, are you aware of anything that these undoubtably better scans offer in a clinical setting? How will they change treatment? We would run DTI occasionally when some surgery is planned and the surgeon wants to avoid a particular tract, but this isn't a common occurance.

Interesting, thanks. I didn't realize multiband was so fast on the Skyra - good to know.

The big advantage in principle for neurosurgery is better resolution of crossing fibers and complex branching, for example the lower/lateral projections of the CST to hand area. I'm not sure if any clinical software can take advantage of these scans, yet, though -- so processing would probably need to be done under IRB in researchy software. There's also ongoing work in several groups to better resolve and possibly even characterize tissue microstructure in and around tumors using multi-shell DWI (especially relating to edema).

Here's one small study that compared several high-angle methods to standard DTI: http://www.ncbi.nlm.nih.gov/pubmed/23254805

Walt Schneider's group at Pitt has also done studies in neurosurgery planning and TBI using HARDI-type methods (no cite handy at the moment).

Checked. Siemens 3T Skyra 2x2x2mm 104 directions, multibanded 6:32. Research scan.

Meta: what a weird and distracting page. From the top there is a menu, logo, menu, breadcrumbs (I guess?), and then content. From left to right there is content, stuff, stuff and more stuff. Only 25% of the page (excluding the header!) is content. It's not even clear to me why I should be interested in looking at the other three columns.

I tried widening the screen across both my monitors. It looked... interesting. http://i.imgur.com/MoKakDW.jpg

I mean, in fairness, if you have a Window that wide, what do you expect to see? It would normally either be blank space, or ridiculously stretched out text. I guess you could split the text into multiple columns like newspapers do, but I haven't seen many websites do that.

"To achieve that kind of speed and resolution, the [Connectom] machine draws 22 megawatts of power"

Can that be right? The average large US hospital uses only 2.1MWe on average.

Source: Table H6 of https://www.eia.gov/consumption/commercial/reports/2007/larg... (57 billion kWhe annually for 3,040 large hospitals)

22MW is comparable to the electricity consumed by the worlds top supercomputer, including cooling system.

It seems they are using a large high performance compute cluster, and not a "branded" supercomputer like the Blue Gene/Q.

http://www.humanconnectome.org/documentation/HCP-pipelines/m...

http://chpc.wustl.edu/who.html

I personally don't see any difference between a large enough computing cluster and a supercomputer.

Marketing and branding. The BlueGene/Q installation is sold as a package.

That's got to be a pretty nasty sequence to have run on you. A high res diffusion scan on a stock scanner can be heard a long way off and I've heard people 3 floors up complain of the noise (although that individual was quite susceptible to noise). This sequence must be rough on participant's ears.

This is one of the greatest recent resources for human neuroscience. This project is giving hundreds (and soon thousands) of subjects' worth of structural, diffusion and functional datasets. This is a huge step toward developing community consensus on how to model human neuroimaging data, which is hindered by the tendency of most labs to keep their data.

The datasets are also great quality, preprocessed and aligned.

Here is a link to the project's website:

https://www.humanconnectome.org/

Assuming development continues as today we should have the first (synaptic level) human connectome ready around 2035. Should be interesting.

What a weird, Hacker-News specific URL. Here's a normal one: http://spectrum.ieee.org/the-human-os/biomedical/imaging/bra...

Thanks, we updated the submission link.

It's not changed for me. I get this url on the HN link:

    spectrum.ieee.org/the-human-os/biomedical/imaging/brain-scanning-just-got-very-good-and-very-unsettling?utm_content=buffera9ad6&utm_medium=social&utm_source=linkedin.com&utm_campaign=buffer#.V3RLFDgLxrs.hackernews

Reply on news.ycombinator.com