Facebook launches MRI research with NYU Langone

Most of us might consider Facebook the social network of choice for suburban moms and conspiracy theorists, but the company did not hesitate to diversify to become a lot more than an app on our phones, even if it’s the last thing we want. Here’s an example: Earlier today, Facebook started a business blog post describing his latest adventure, this time in the wild world of medicine.

As the article explains, the company’s AIresearch wing—Called FAIR — has spent the past two years quietly working alongside professionals at NYU Langone Health Center to create what they call fastMRI: an algorithm that promises to reduce the lengthy process people typically go through when getting into an MRI machine. They just need a few photos of your bones to do this.

Okay, not all of your bones – at least not yet. The first round of research for the fastMRI program, at least so far, is based exclusively on a open-source library images from a series of knee MRIs that NYU helpfully donated for the good of the project. By training a machine learning algorithm on these knees, the team were able to create an algorithm capable of estimating an accurate MRI image using a quarter of the data from your typical MRI machine. take in when creating a crystal clear image of your bones or your brain or whatever. Or in other words: because an algorithm does the heavy lifting here, you spend less time having your photo taken while trapped in a weird, noisy metal tube.

You see, the main reason a given session in an average MRI machine can last is More than an hour comes down to how these machines work in the first place, which is … kinda complicated explain. In short: if a machine is, for example, scanning a certain person’s head (or the brain, as the case may be), that means applying super strong magnetic forces to a person’s head, shooting that head with a radio current and then forming a composite image based on the behavior of how many protons of God knows how many in that person’s head once they’ve been subjected to those kinds of signals. It turns out that the signals emitted by these protons can be very weak, which means that all of the mess may have to be repeated over and over again in order to form a crystalline composite.

Use AI to reduce the time it takes to get that final image is not a new idea by far, and I’ll be the first to admit that sounds like a great idea – until you remember that Facebook is one of the names behind this particular project. It is a company whose insane growth rate largely relies on collecting our data, aggregating that data, and then transferring it to third parties such as major advertisers or federal agencies. And that doesn’t include, you know, all these massive The data violations the company continues to be at the center of.

And just like Facebook’s ambitions as a platform, its data sources have also rapidly diversified: Facebook doesn’t only know what we are doing on its platform or on Instagram, it also knows what we buy, where we buy it, well, a ton other things, thanks to its multitude of partnerships in a buffet of industries, including “surprise surprise”large pharma and medicine.

This year alone, the platform made a noticeable push courting the advertising dollars of major medical brands, and this has been job, in part thanks to the pharmaceutical data that we have freely to abandon about ourselves online. And if that part of our medical history is used for targeting, then, well, there’s nothing stopping Facebook from doing the same with any other type of medical dataset, even one that comes from our literal bones and organs. . In general, these types of details are meant to be covered by a legislature like HIPAA, but like we covered before, the lines between what can and cannot be monetized become a bit blurry depending on whether that data comes from a doctor or a tech company. And in the case of something like fastMRI – or something like The alphabet is really– we have a partnership between public health and privatized technology, which means that HIPAA might not protect these MRI scans as much as we hope.

Facebook undoubtedly saw some of the potential discomfort before posting this blog post, as the company sneaked in this little warning towards the middle:

(The fastMRI data used in the project, including the scans used for the study, is from the open-source dataset that NYU Langone created in 2018. Before opening the data, NYU Langone made sure that all scans were de-identified, and no patient information was available to reviewers or researchers working on the fastMRI project. No Facebook user data was provided for the creation of the fastMRI dataset.)

Okay, it looks like these bone scans aren’t being used for tracking and targeting – at least, not yet – but the team’s own blog post makes it look like the fastMRI project isn’t going to go down. stop with images of strangers knees. “Today’s clinical study is an important step forward, but there is a lot more to come. Next, researchers at Facebook AI and NYU Langone want to show that rapid MRI works just as well with other vital organs, such as the brain, ”Facebook wrote.

Even though we’re going to scrap the whole ‘my bones are used against me’ narrative (which, I admit, is more speculative than I would prefer), there are still a ton of reasons you shouldn’t. want this company anywhere near your medical data. It’s a company that repeatedly demonstrates that it will put its profit margin ahead of the safety of its users, no matter how hard it tries to pretend otherwise. Hell, on the very day that Facebook published this AI research, reports revealed that the company was always with proof about his role in the wave of genocides in Myanmar in 2017. Here in the States, society struggled continuously to take action against antivaxx groups that have already caused at least the death of a child. And, of course, there is all the so-called goat murder thing.

I’m not saying Facebook’s medical research isn’t really going to help anyone, but I’m saying rapid MRI might be less tricky if run by absolutely any other company.