BRATS Africa: Building Inclusive AI in Radiology
Bibliographic record
Abstract
Intro-From the RSNA, this is the Radiology Artificial Intelligence Podcast.My name is Paul Yi, and I'm a radiologist and co-host of the podcast.And my name is Ali Tejani, and I'm a radiologist and co-host of the podcast.Each month, we dive into the hottest topics in radiology AI and talk with leading experts, thought leaders, and movers and shakers in the field. Dr. Ali Tejani-Welcome back to theRadiology AI Podcast.Paul, you know, it's been a while since we've been together to record one of these episodes.How have you been?I feel like I haven't seen you in a while.Dr. Ali Tejani-You're joining us from somewhere that's not close by.Where are you today?And what are you doing?So currently I'm in Lagos, even though I still maintain my appointment at the University of Pennsylvania.The idea, the main reason why I'm currently in Lagos right now is the fact that there's a lot of advanced technologies that have been developed out there, and because of the infrastructural set of healthcare and all the women in the country, they are not able to laterally adopt those technologies.So we created my lab in 2022 with the assistance of Dr. Ana Soto, who is on the call with us today, to adapt those advanced technologies to meet resource constraint settings in Africa and other resource constraint settings in other parts of the world.Dr. Ali Tejani-I know we're going to have a huge conversation that we're, I'm sure all of us will learn quite a bit more about that as well.So thank you for joining us.And Udunna, where were you recently?I feel like you're probably traveling as well.Dr. Udunna Anazodo-Yeah.So I was where Marouf is right now.I was actually at my lab at the medical intelligence lab in Lagos.And also I visited a couple of sites where we're starting to do some research in Nigeria.I was at the region's healthcare center in Ower, which is southeastern part of Nigeria.So Lagos, for people that don't know Nigeria, Lagos is on the southwest.I think that's where most people that know Nigeria can pinpoint where the country is.And if you move five or six hours eastward, there is a town called Ower, and there is a fantastic private center, but actually on a multi-specialist hospital.And they have an MRI scanner now, which we're trying to optimize and essentially make sure that I could produce very similar type of imaging quality as we do here in the west.And I also traveled up north, my first time being up north in the country, to a town called Damaturu, which is in Yobesti.And that town right at the border of Chad and Cameroon, so it's at the border town.And it's also a town that was heavily aXected by the Boko Haram insurgency that happened a few years ago.So we're at the hospital that started a very fantastic dementia research study, and there we're trying to, again, see how well we can do imaging within the constraints that they have.So I've been moving around Nigeria, and I think we're going to talk about some of the work we're doing, especially the BRATS Africa project.Dr. Paul Yi-Yeah, so why don't we just jump right in?So you mentioned BRATS, which is short for brain tumor segmentation.And for many of our listeners, they know that this is an annual brain tumor segmentation data science challenge.It's been held for over a decade, if my recollection serves me correctly.Can you tell us more about that?What is BRATS?How did it get started?And how did you all get involved?I think you mentioned that there is an MRI scanner in the city that you had mentioned, which implies that maybe these things aren't so common.Tell us a story.Dr. Udunna Anazodo-Right.So BRATS, like you said, is a brain tumor segmentation challenge.It's a challenge that's been run by the Medical Image Computing and Computer
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.016 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.002 | 0.005 |
| Scholarly communication | 0.007 | 0.013 |
| Open science | 0.003 | 0.017 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.036 | 0.013 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".