Just as a general note, these images contain personally identifying information about the patient. This includes not just the patient's name, but also his age, address, phone number, and an extremely limited health history.
For the record, i wanted to post my brain online for many reasons. First of all, I did it for science of course. It'd be great for more of this data to be available online for people to study and work with.
There are many anonymized datasets available on Radiopaedia. http://radiopaedia.org/ I've spent quite a bit of time studying neurological conditions on that site and on wikipedia. Those are more useful, but anonymized. I'm not sure that there are many healthy MRI datasets available there, which makes it more difficult to statistically model MRI data.
I'm hoping to use Nipy and Dipy to create visualizations of my brain. In particular, i'm hoping to recreate some of the Diffusion Tensor Imaging work in this talk by Ariel Rokem: https://www.youtube.com/watch?v=HdrSZtB0uX0. However, i'm not sure if my DT images are high enough resolution.
I also did it for philosophical reasons. I believe that all digitized information past and present will be inevitably become known & public post facto. It's the nature of information & knowledge in our technological world to distribute itself until the world converges on total information omnipresence. So if, in the end, all [digitized] information is known, then there is no point in hiding information. This is more of a philosophical viewpoint though. obviously it's impossible for all information to converge in this way.
But this is true even of our own minds. And if our mind can be digitized from our brain, this also has interesting security implications. If you've seen the show stitchers, then you know what i mean.
Under no circumstances was this done in the name of art.
Under no circumstances was this done in the name of art.
Out of curiosity, why point this out? It's obvious that you didn't do it for a number of reasons. Hundreds, in fact. So why art?
I also did it for philosophical reasons. I believe that all digitized information past and present will be inevitably become known & public post facto. It's the nature of information & knowledge in our technological world to distribute itself until the world converges on total information omnipresence. So if, in the end, all [digitized] information is known, then there is no point in hiding information. This is more of a philosophical viewpoint though. obviously it's impossible for all information to converge in this way.
This would be the end of freedom. I'll go into the reasons if needed, but the discussion has happened a few times before. Suffice to say, if you're not allowed to have privacy, then you're not allowed to be free.
Freedom is vague, here, but I think what you want will still be possible (at least theoretically) given causality and physical distance.
Deeper freedom (that is, infinite potential) is inherent to being itself, but if what you're looking for is something more like diversity or novelty, then yes, that's contingent on perceptual subject-object duality.
Causality and physics is also what makes GP's wish near-certainty given continuing technological progress. Hiding information is going to get only harder.
Thanks for posting your scans. As noted in the issue I just made (I am @maedoc on Github), if you add the diffusion imaging scans we can virtualize your brain [1] and simulate it [2].
What strength MRI was used for the scans? 1.5T, 3T, 7T?
I doubt the scans really high enough resolution for a meaningful simulation --- I don't believe MRI scans can capture the complete connectome of a single human brain, let alone any subcellular data.
e.g. the Human Connectome Project is still working on capturing the macroscale connectome, that is, high level structures only. Recording the microscale connectome, that is, the neuronal structure, currently requires destroying the brain you're analysing.
As the research engineer on the Virtual Brain simulator to which I linked, which we apply to improve e.g. epilepsy diagnosis, I can say that those scans do have enough resolution to capture global properties of anatomy and dynamics of brain networks.
Yes, but that's just large scale structures. This wouldn't be simulating any of the features which most people find interesting about brains, i.e. the personality. It's important to make that clear. This is not a brain upload.
Personality would be expressed as interperson variability in certain parameters or connectivity of the brain model. It's not at all mystical or hiding in the subcellular structure.
Brain upload is a sci-fi invention; using a T1, DWI, fMRI, EEG and behavioral data to build a realistic model of a subject's brain is today's neuroscience.
Brain upload is a sci-fi invention; using a T1, DWI, fMRI, EEG and behavioral data to build a realistic model of a subject's brain is today's neuroscience.
I will echo this point. I don't think I've met anyone in the field of neuroimaging who sees mind uploading as plausible. There's simply too much noise in the data for that sort of thing. Even simplified simulations such as the Human Brain Project require so much computing power that modeling a single mind is prohibitive; I can't imagine trying to perform whole-brain emulation accurately and practically without incurring more cost than it's worth. Wetware is good enough for me.
> (1) Even simplified simulations such as the Human Brain Project require so much computing power that (2) modeling a single mind is prohibitive
This step from (1) to (2) is a common assumption that most people are unwilling to reconsider, but it's equally possible to reject bottom-up modeling and ask what are the high-level constraints on brain activity that shape and direct it. This is what we do with the data I mentioned (T1, EEG, etc).
You need at least 10 picometer percision to go down to the cell level (1.5L / 86 billion neurons), MRIs are 1mm according to the talk dcunit3d posted [0].
Yikes, please don't step anywhere near anyone else's medical data. That kind of "philosophy" is pretty much an imperative to violate patient and research subject privacy (and yes that is something to be protected).
You say there is no point in hiding information if it will be known eventually but "eventually" is a key point.
Almost all information, especially personally identifiable information has a temporal component to it.
For example, that record you posted opens the door for someone to find your (dcunit3d's) current address.
If you recently pissed someone off on the internet (even under a different username), knowing your current address would enable that someone to send a litany of irritating, potentially damaging stuff your way.
On the other hand, if one of your records from five years ago ends up public, the address listed on that record is most likely not where you are living now so it is of little immediate use to a would be attacker.
Another fairly well know example is the AshleyMadison hack.
In that case, the recency of a given user's record on the site made the difference in whether the hackers were able to successfully extort money from the victim.
Also, I wanted my brain to be online in a widely distributed, public manner. In case of robot apocalypse of course. Hopefully, a quantum computer in the future would have low-bandwidth write access to my brain in the past. Quantum mechanics. Then I would be one of the earliest widely-distributed, targetable maps of a human brain on the public internet. And the people of the future could avert a worldwide AI disaster by pinging memories directly into layers of my brain. A la Steins;Gate, if haven't seen it, you should.
Humble neuroinformatics lab knave here. While it's nice that you've made your data available, you might want to reconsider including the PHI. Many researchers are not interested in your phone number / where you live past the recruitment stage, and even then that information is kept confidential in accordance with IRB policies and HIPAA. The only data that a researcher might want to know is whether there are any phenotypes associated with your data (e.g., autism, ADHD, other stable traits, etc) or, in the case of task-related data, when certain blocks or events started/stopped and for how long. Even with phenotyping data, we are very cautious about how we go about sharing since we don't want to violate anyone's privacy (see here: http://fcon_1000.projects.nitrc.org/indi/enhanced/sharing.ht...). This is very important for individuals with conditions that are stigmatized in society- we don't want to make someone's life worse by outing them as being autistic, bipolar or depressed for instance.
In light of what I've said above, I will also express my disagreement with your philosophy. I do not believe that is natural for information to converge to a single point- in fact, many corporations rely on information asymmetries to gain an upper-hand in the market. The internet may reduce such information asymmetry, but it is unlikely that it will eliminate it completely (unless humanity turns into the Borg and creates a hive mind). Indeed so long as human thought is decentralized with separate minds in separate bodies, information asymmetry will remain the default.
Moving along, I want to note that a chief goal of science is to seek truths that are generalizable to the population at large. A single DTI series isn't really useful for achieving that goal since you can't determine how it relates to scans from a number of other participants(i.e., a 1st-level analysis is not a 2nd-level analysis; a case study is not a wider truth about cognition). In that sense, it might be better to see if there is a databank willing to accept your scan and donate your data to that so that you could be part of a broader sample.
More pragmatically, you might also want to consider using the NifTI format for your data, since this is the format that researchers use when processing data in various neuroimaging suites (i.e., Freesurfer, AFNI, SPM, FSL, etc). You may also want to consider organizing your data according to the Brain Imaging Data Structure standard, as many datasets are moving towards this (http://bids.neuroimaging.io/) and software tools are being written to take advantage of this structure.
Finally:
Under no circumstances was this done in the name of art.
The original poster (being savvy enough to use GitHub, etc.) was probably aware of this but didn't think it was worth his time to carefully edit out that info.
It's too bad we don't have a universal way of sanitizing all meta data from JPEGs, PDFs, Word files, and everything else at the click of a button. Something that's really simple and quick to use.
I'm aware that there are lots of separate little tools and hacks to deal with sanitizing or anonymizing, but I wish there were either a standalone tool that dealt with every kind of common file, or a "Sanitize" button that was a standard feature in every viewer, editor, or browser (when uploading for example).
You can export anonymised files from Osirix very easily. You can also just anonymise the fields you want to easily. I'm yet to see a dicom viewer as good as Osirix - it's simple to use and has every feature I've ever wanted. How the paid viewers get it so wrong is beyond me.
it was in the name of science, of course! i'm pleasantly surprised to see some of the discussions on here, concerning copyright and health data ownership. =]
i was aware of the privacy implications. i briefly looked at the data in osirix that i would be exposing and decided that it was worth it.
I'm pretty sure that he's legally allowed to release his own PHI and PII. Since the README includes "I open sourced my brain. Here it is," it's from "dcunited001" and the scans are for David Conner, I'm pretty sure this is the person in question releasing his own information.
I don't think people are questioning the right for him to release his MRI scans. I think we are wondering if the patient is aware of how much of his private info is being released in the scan.
Most definitely. Using the pydicom library, you can view all of the metadata stored along with the images. For instance, the patient's telephone number is listed with a 540 area code.
I have not and will not peek at the DICOM, but anyone else thinking about doing something like this should heed the preceding comments and be aware that stripping PHI headers is critical to deidentifying medical imaging data. Also, the imaging data itself can also be identifying — best practice for sharing deidentified neuroimaging data involves face-stripping.
Last year an MRI scan was made of my brain. This was part of a promotion research at the university. All I got back was a lousy screenshot of my brain. ;-)
No but seriously - it was a big shock to see my brain in a picture. It is a really weird experience, similar when I saw myself for the first time on film. Maybe this is a normal experience for the current generation, but I hadn't seen myself on film before age 15. You have an image how you move, what you look like, but then you see this on film, and it's a total shocker. All these small movements you make, typical for you, and everybody around you knows them, except for you. So everybody else sees nothing strange when viewing that movie, except you.
I've seen many MRI scans in films and tv series, although I can't remember one since then. From these I have a general picture of what a brain looks like. I've seen plastic 3D brain models. And then I see my brain and it's so different. It's clearly me, no doubt, but still... Totally weird!!!
How do you feel like a low resolution 3D scan of what someone told you was the inside of your head is clearly you? I'd feel like they could've shown me anyone's MRI and I'd have no way of telling them apart.
I got one screenshot with three images in it. The (vertical) profile image is clearly me. The other two (vertical frontal and horizontal at eye level) that could be anyone.
Back in December of 2007 a nodule was detected on my lung. Not knowing what to think of it, I posted the images on my blog and recruited other med students to comment on the images. Here's the Way Back Machine link to the blog post.
Not to comment on your situation specifically, just some interesting related information:
The national lung screening trial [1] which had on the order of ~25k CT scan patients and ~25k radiograph patients found 96.4% and 94.5% of the positive screenings to be false positives. This screening trial has been a large motivator for computer-aided detection, diagnosis, and characterization of lung CT scans (although work on this goes back to the 80s and 90s).
If you look at the fleischner guidelines [2], the key factors in determining follow up are patient background and nodule size and growth rate. There are secondary diagnostic characteristics that are correlated with malignancy as well, e.g. spiculation and texture patterns.
Ended up being benign. I probably went a bit overboard, but did a pet scan, nothing lit up. Continued to do CT scans to monitor for growth, and it stayed the same. I was just so scared..
Was wondering, in the general, is the patient always the copyright owner of this kind of data? Can they relicense it, publish it, do anything they want with it?
It seems like that would be the case, but there has been talk about patenting parts of the DNA in the past, and legally sometimes things work in un-expected ways.
Whoa, hold on there. HIPAA generally gives patients the right to access their medical records [1], but I don't think it says anything about whether they "own" it in the sense that they get to edit the data to their liking (I don't believe they do, in the US), and it certainly doesn't say anything about copyright.
Most of the HN crowd (and myself) would probably say that the ideal situation is for patients to own copyright and other rights in medical data generated about them, but I don't think this issue has actually been settled, either in a cultural or a legal sense [2].
Yeah that's why I asked. On first thought it was "well yeah, it is their brain of course they own it". But then I could think of scenarios where imaging centers, doctors, hospitals, equipment manufacturers could get wierd about it.
I found a case in Canada, for example, about a hospital asserting ownership rights to a biopsy tissue sample.
Then remembered the DNA patentability case not too long ago as well:
I would argue that patients owning the copyright on their health data is actually a bad thing. It would make a lot of research much more difficult if they had to get rights to publish patient data even after de-indentification. We NEED that data to drive the field of medicine forward and as long as the patient's privacy is not compromised, it would make sense for the hospitals to own the data.
I can see your point, but almost all clinical research already involves a consent process where the patient is told the benefits and risks of participating; signing away a license to a subset of their medical data could simply be included in the consent.
There are some regrettable examples of cells and tissue being taken from patient specimens without their knowledge, some of which give birth to entire fields of research, and then it is later unclear whether data that could re-identify the patient's family (like genomic data) can be published [1]. It can be really hard to fully de-identify imaging and molecular data.
> I can see your point, but almost all clinical research already involves a consent process where the patient is told the benefits and risks of participating; signing away a license to a subset of their medical data could simply be included in the consent.
You don't need to get consent if the PHI is de-indentified and the data was obtained through a regular course of treatment (not a study). This is really important because hospitals can publish anonymized case reviews for rare disorders or perform chart reviews on patient outcomes (drug X produced Y outcome in Z patients with [some condition]). I read through a lot of these papers on a daily basis and can't imagine the state of medical literature if consent was needed for publishing this clinical data. Retrospective chart reviews are typically exempt from human subject regulations.
From that perspective it seems equally absurd that a person wouldn't automatically own the copyright on a photo of themselves, and yet they don't - the photographer does. It seems plausible that the person who creates the physical record owns the copyright, regardless of the subject. (I don't disagree with you. Just a counterpoint.)
Someone further down the thread linked a piece about x-rays, but in general I'd suspect that an MRI or other visual produced by medical scanning technology would not be copyrightable in the first place; copyright does require a certain amount of creative effort in the production of the copyrighted work. The important case here, if you want to look up the history, is Feist v. Rural Telephone, which dealt with the (non-)copyrightability of a list of telephone numbers.
I'm not sure how definitive that guide is. For what it's worth, all of my medical textbooks are very careful to include licensing and attribution information for all of their medical images, including X-rays. So even if there's not much case law on it, publishers seem to be erring on the side of caution.
Interesting. I saw the MIT license and was wondering (mostly in general, not about this particular case) how that would work. I understand to re-license something, one must hold copyright ownership, and unfortunately, I could imagine imagining companies or even hospitals getting a bit strange about that.
hmmm, good point. i think the copyright issue would be fairly complicated here ... and you'd need a good lawyer basically. it'd likely hinge on how you constructed the argument.
Of note: if the scan is not done for diagnostic purposes, i.e. if no doctor looks at your images to do a diagnosis, then it is not considered medical data at all. Just like taking a photograph of yourself.
(I was a little let down I couldn't follow the visual nerves, but as they cross there is no clear direction of diffusion that could be imaged I guess...)
Yes. For my girlfriend's birthday I grabbed her MRI scans, extracted the pial surface as a mesh, 3D printed the mesh, made a silicone mold from the print, and cast her brain in chocolate so she could eat her own brain, thus becoming an auto-zombie.
I did this too, but it took much longer than the simple steps you've outlined would seemingly take. I spent at least an hour on the whole process, which is more than I'd prefer for something so straightforward. I blame the early tech of 3d printing for not being as swift as I'd hoped.
(30,380 cases. 187,945 images.)
its a teaching/sharing site for radiologists where they can send the data from their PACS.
this is the free public site which includes a basic web based dicom viewer to zoom/pan/window level etc
Here's a lot of data from Russ Poldrack, who scanned his brain and collected behavioral and metabolic data from himself very regularly over the course of a year: http://myconnectome.org/wp/data-sharing/
All the MRI scans of my brain I've gotten ("structural" scans taken as part of fMRI studies) have been sufficient to get a pretty decent volume rendering of my head. Some of them cut off about half way through the jaw, but that's another issue entirely.
The research scans use 3D T1 (and sometimes T2) scans for mapping the functional data onto a brain region. They have to put the low res functional scans onto something that has some degree of resolution. Functional scans are acquired with temporal resolution in mind - the change in signal after activation is something like 6 seconds, and to accurately sample this you want a whole brain acquisition in much less time that that. We typically acquire whole brain scans in something like 3 seconds. New techniques such as multiband imaging are acquiring whole brain scans in 600-700ms or so. This is pretty fast.
Clinical scans do use 3D techniques but tend to go for 2D scans (which are a slice, then a gap of 10-30% usually of slice thickness) then a slice.
2D scans are faster, have much better in plane resolution (lower through plane as slices are fatter and have gaps between them). Clinical scans are also one hell of a lot better for identifying things as numerous image weighting a are obtained. MR tech here.
What's the resolution of the whole brain scan you deal with in research these day? (ie the cubic size of the voxel when acquired in 3s? What about in 600ms)
I have to keep myself up-to-date, it's been a while since I played with that!
Look for research groups in your local institutions and send an email to a couple grad students asking about studies they or their colleagues are running in the near future. A researcher may give you the scan if you ask nicely and say it's for personal interest, plus you'll be compensated for your time /and/ simultaneously help with the advancement science :)
To expand on Deniz' answer, Klarismo (YC S15) does a full body MRI, including a high resolution scan of your head for $249, available in SF and LA at the moment. We extract your brain, segment out gray and white matter and measure your brain volume. In addition we segment, measure and visualize visceral fat, subcutaneous fat and major muscle groups.
This site has thousands of anonymized MRI and CT images of normal and abnormal scans, with clinical history in many cases. For example, here are annotated scans of normal anatomy:
One challenge with writing web DICOM viewers is the lack of full 16 bit image support in browsers. This requires writing custom code to properly handle window / level on 16-bit medical images.
Wow. This goes to show that people in field X often have no idea what's interesting to people outside their own field. My PhD is in cognitive neuroscience, and so I have many dozens of MRIs of myself done for test or calibration purposes, done using all sorts of different machines, sequences, etc. It's never occurred to me that these images would be of interest to anyone but me.
I suspect this is true for pretty much anyone in any field.
If we get better at this and are able to create detailed enough scans of a brain maybe we can recreate that person some day, including memories and everything.
As kendallpark pointed out this is highly unlikely. I upvoted you because it's an interesting "hmmm... what would it take?" The spatial resolution of an fMRI is high compared to something like EEG, but the time resolution is much slower. Even the spatial resolution is on the order of cubic mm at the absolute best. So, we are no where close to being able to scan the activity of every one of the ~100 billion neurons in a human brain and it would 7+ orders of magnitude better spatial recognition and associated processing capabilities. So, the magnitude of your "if" is large enough that it's on the level of Kurzweil Crazy, but cool to think about regardless.
I know nothing about neurology, but my money would be on using a different mechanism to extract all the data, maybe something like fake neurons that dump all the information via WiFi, a MITM attack for the brain itself if you like.
With our current understanding, such tasks are comparable to trying to break the speed of light. However, we could probably get incredibly close without any internal data extraction. Our personalities, while highly complex, can be approximated into some simple traits (e.g. 74%
'passive-aggressive', 47% empathy) and replication of our knowledge isn't strictly necessary. Find 10,000 interests and create a int_8 each of how much someone is into it. Et cetra.
I don't think we need to scan 100% of the brain in molecular detail in order to read & preserve the information. It should be possible to observe the circuitry at a higher level and recreate an approximation -- though it'd be just that, an approximation. Additionally, much of the information could be compressed to a much more lightweight format than digitizing the physical neural networks of the brain.
Digitizing a static physical model of the brain is problematic in itself. Many regions and structures in the brain are likely structurally dynamic. That is, the neural circuitry could change structure from day to day and this is definitely true over longer timescales. So it'd be much easier to read the data from the brain into another format rather than emulate the physical structure of the hardware.
I've thought a bit about information could be read from MRI scans such as these. In the absence of higher resolution data, I don't think there could any memories retrieved. I don't know the low-level details of how MRI data is encoded, but I do know it is recorded in slices. To retrieve memories, you'd need to read 3D regions from the brain. And it'd need to be much more high-resolution.
It's freeware (though not open source) and an excellent and fast image viewer. You'll have to also download the plugins to view the Dicom images (the MRI images). Fortunately all plugins for IrfanView can be downloaded as a single file as explained at the site above.
As other have already mentioned, the MRI images are full of personal meta data (which can be shown by clicking "info" in IrfanView). I assume that the OP must have been aware of this.
DWV [1] (open source, Flash/Java-less webapp) Seems to still be a WIP, but it lets you load the files from the repo (to localstorage), despite throwing errors left and right. ;)
Not quiet the same, of course, but this reminds me of the game Soma, where [SPOILERS!] the protagonist has his brain fully mapped and "uploaded" as part of a treatment, and his scans are used as a template for A.I. research in the future, leading him to be "reborn" a century later in a different body.
For those who are interested, the human connectome project is in the process of acquiring and publicly releasing a database of 1000+ MRIs collected from normal subjects:
