Play with us! fibr.dev, a citizen science project for MRI quality control

Feel like playing a swipe app that doesn’t involve dating? Do you want to help scientists understand pediatric mental health? Try our new citizen science app at https://fibr.dev.

What is this thing?

We developed a citizen science web app to help measure the quality of brain MRI scans . To play, simply log in to https://fibr.dev and swipe left if the brain scan is “bad” or swipe right if the brain scan is “good” (don’t worry, we describe what “bad” and “good” mean in a short tutorial).

Why we’re doing this?

Innovations in medical imaging are helping researchers understand the developing human brain. There is growing hope that these technologies will lead to clinical tools that can help diagnose and treat diseases, mental illness, and learning disorders. But to achieve this dream, we need massive MRI datasets to capture the broad spectrum of brain development. So the Child Mind Institute has launched the Healthy Brain Network (https://healthybrainnetwork.org/), a landmark mental health study that is collecting MRI images and other assessment data from 10,000 New York City area children. We have developed algorithms that analyze these images to measure the integrity of connections between different parts of the brain. But they aren’t perfect and sometimes do a bad job, especially if the child moved a lot in the scanner (kids have a hard time staying still sometimes). We need humans to visually inspect the quality of the output. Computers can struggle with this sort of thing but humans are great at it! That’s why we need you. We need you to help researchers learn about pediatric mental health and learning disorders.

How you can help

Log in to https://fibr.dev. Classify these images as “pass” or “fail” by swiping right or left. You’ll get points for each sample you swipe and you can unlock fun animal badges based on the number of points you have. Go unlock them all so you can brag about being a unicorn!

4 Likes

Hi,

First thoughts on the game:

  1. I’d definitely like a more interactive training period, as opposed to just reading instructions for 10 min (yes, I know, this is really picky, but most people are picky/low commitment like this). Specifically I’d prefer something like: having a set of example scans with comments on why it is good or bad, and then after each scan/reason, making us swipe a couple to make sure we understand what you are describing.

1a. Check out Sébastien seung’s labs “EyeWire” game to get some ideas. I particularly enjoyed and learned a lot more from their tutorial format. That game has a very elaborate scheme for getting people trained which involves an interactive (rather than purely reading) tutorial, and you have to have above a certain accuracy to be allowed to grade samples that have not been graded by anyone else before. (your accuracy is based on comparison to either expert verified answers, or crowd sourced answers)

That is my main criticism. I think on top of increasing retention, it would also make the answers you get more accurate.

Otherwise, love the idea of crowdsourcing science so scientists can do something less mindless.

Thanks for the feedback!