AI can spot 'invisible' brain abnormalities in  patients - video Dailymotion

The Manila Times

AI can spot 'invisible' brain abnormalities in  patients  Scientists in London are using AI to detect previously 'invisible' brain abnormalities — too subtle for standard detection methods — that cause epilepsy.  The tool, called MELD Graph, spots around two thirds of the tiny abnormalities that human radiologists often miss when studying a patient's MRI scan.  The team have spent 10 years developing MELD Graph which was trained on MRI data from more than 700 people with focal cortical dysplasia (FCDs), a major cause of epilepsy. FCDs can be subtle and difficult to see with the human eye and up to half of these lesions are missed by radiologists.  The system highlights what it thinks might be an abnormality and also explains why. The analysis can then be checked by a radiologist, speeding up a diagnosis and access to surgery.  REUTERS / KING'S COLLEGE LONDON / UCL  Subscribe to The Manila Times Channel - https://tmt.ph/YTSubscribe  Visit our website at https://www.manilatimes.net  Follow us:  Facebook - https://tmt.ph/facebook  Instagram - https://tmt.ph/instagram  Twitter - https://tmt.ph/twitter  DailyMotion - https://tmt.ph/dailymotion   Subscribe to our Digital Edition - https://tmt.ph/digital   Check out our Podcasts:  Spotify - https://tmt.ph/spotify  Apple Podcasts - https://tmt.ph/applepodcasts  Amazon Music - https://tmt.ph/amazonmusic  Deezer: https://tmt.ph/deezer Stitcher: https://tmt.ph/stitcher  Tune In: https://tmt.ph/tunein  Soundcloud: https://tmt.ph/soundcloud   #TheManilaTimes #tmtnews #artificialintelligence #epilepsy

Transcript

00:00So you can see that we've got a gray matter with a high polar intensity and I was trying

00:17to find if I could see it myself on the scan.

00:18You can see that there aren't any lesions elsewhere using CKMC?

00:19Yes.

00:20That's what I was going to say.

00:21Pretty cool.

00:22What causes the lesions?

00:23It's just slightly concerning because they couldn't...

00:30Do you think any of the lesion masks, because I thought, I remember in Mel 1, people drew

00:46lesion masks around the transplantar sinus.

00:50So about 1 in 100 people suffer from epilepsy, and in about a third of these, the anti-seizure

00:56medications don't help to control the seizure.

00:59For some of those patients, a brain abnormality, so a part of the brain that's not formed correctly,

01:05is causing these seizures, and if you can find it, then you can operate and potentially

01:10cure the seizures that way.

01:11I think actually you can really see the transmantle on the T1.

01:12Yes.

01:13A bit more clearly.

01:14Yeah.

01:15Just there.

01:16It's here.

01:17It's part of the vision.

01:18It's a brain.

01:19Yeah.

01:20Yeah.

01:21It looks a bit bigger.

01:22Well, I think the prediction...

01:26So we've made an AI tool to help diagnose hidden brain abnormalities in children and

01:31adults with epilepsy, and the crucial thing is if you can find the brain abnormality,

01:36you can offer surgery which might cure the seizures, and what our AI tool can do is find

01:42two-thirds of those abnormalities that doctors might miss, and that's really important because

01:47it means that more children can get the best possible form of treatment.

01:51It's not size that matters.

01:55The smaller the brain is, yeah.

01:59Yeah, that's true.

02:01So we see them slightly bigger than they are.

02:08So one of these gone out, and sometimes there's almost no noticeable...

02:13You know, if you talk to them on the phone or on Zoom, you wouldn't notice.

02:17Obviously, I think some of the motor stuff you would notice because they're...

02:20It's called MELDgraph, and I can talk about MELD and graph.

02:23So MELD is the Multicenter Epilepsy Lesion Detection Project.

02:27So this is 23 hospitals around the world that have found examples where they found these

02:32brain abnormalities and shared them with us, and that made a large data set of 700 of them.

02:36That's more than most radiologists would see in their lifetime.

02:39What we're then able to do is take the graph bit, which is an AI algorithm,

02:43and show it lots of these examples so that it can learn what types of things to look for

02:48and how to find these brain abnormalities.

02:50And then the last bit is what we've then shared, is it can take a new scan,

02:54point out where it thinks the abnormality is on that scan, and also, really importantly,

02:59how confident it is, 9 out of 10, 10 out of 10 that it's a brain abnormality,

03:04but also what it's using to make that decision.

03:07Is the brain a bit too thick in this area? Is it a bit brighter than it's expecting?

03:11Then a radiologist can go back to that report and have a look at the scan and say,

03:15yes, I think this is causing the seizures. Maybe we can discuss surgery now.

03:20A translantal sign, just there, going to the ventricles, which may be the FCD.

03:27Can't see it, but it's useful having the AI point out where it is,

03:32which is, I think, what it's pointing at.

03:34And then we can also have a look at the T1 and see if we can see the blurring.

03:40That one, I'm really not sure I can see.

03:50I think it's important to put into context how difficult these can be to find.

03:55An MRI scan is maybe 10 million pixels, and they're looking for something

04:00that's maybe 100 pixels at its smallest, and that can be very subtle.

04:04For context, an example might be a novel, a 200-page novel.

04:09There's 100,000 words in it, and they're looking for one word

04:12that's slightly the wrong size or slightly the wrong font.

04:14You could see how having an AI to point out potential areas

04:18can really speed up that process and also pick up things they might have missed.

04:41Perhaps I can give a case example of an 8-year-old girl

04:44from scenic Great Ormond Street Hospital.

04:46She'd been having seizures since she was four years old.

04:49The medication wasn't helping to control the seizures,

04:53and so she was referred to Great Ormond Street.

04:55The radiologist looked at the scan.

04:57They thought there was a brain abnormality there, but they couldn't be sure.

05:00What's different now, rather than having to go many more investigations,

05:04many more delays, is they are able to run MELDGRAF on this patient,

05:08and it highlighted an area that they then were able to confirm was the lesion,

05:12and she can now be considered for surgery.

05:16For more information, visit www.ISGlobal.org

05:46For more information, visit www.ISGlobal.org

AI can spot 'invisible' brain abnormalities in patients

Category

Transcript

Recommended