A group of scientists led by University College London has created AI algorithms that is able to find brain abnormalities anomaly in people that have a diagnosis of epilepsy. This approach was implemented in the frames of the MELD program: to create an algorithm, more than a thousand MRI images from twenty-two brain-disorders treatment institutions from different countries were collected. With its help, you can find out where the affected areas of the brain are located in case of a violation of the structure of the cerebral cortex and white matter (fοcal cοrtical dysplasia (FCD)), which is the main cause of chronic non-infectious brain disorder.
How to recognize brain damaged places in patients using AI
The root cause of epileptic disorder is FCD — sectors of the brain that develop atypically. They are the ones that frequently lead to medicine-resistant epileptic condition. As a rule, such a disorder can be cured surgically, but the detection of the affected parts of the brain using MRI is quite problematic since MRI images of the brain can show normal indicators.
To create the algorithms, the scientists obtained a statistical assessment of the features of the cerebral cortex based on tomography images, using approximately 300,000 points throughout the brain to obtain more accurate data. After that, the developers trained AI on the scans received, which doctors called either a normal state of the brain or the presence of anomalies. So, the results showed that the computer was able to detect an atypical condition in 67% of cases (538 individuals participated in the study). When making the diagnosis, doctors conducted analysis with the results of 67% of cases in 178 individuals to be negative – they did not see an anomaly, but the algorithm identified the condition in 63%, which confirms its effectiveness.
Diagnosing abnormal conditions in the brain is extremely crucial, because if doctors can detect them with MRI scans, then operations will lead to recovery.
The developed method can help identify more similar conditions in patients of different ages and allow more patients to have a chance for surgical treatment and further recovery of illness.
Statistics show the following:
- In the United Kingdom, more than half a million people suffer from this disorder. This is even though for most of them medication is not accessible for some reasons, and for twenty-thirty percent of people that way of curing is not suitable as their organisms do not accept such medicine.
- About 1% of the people of the earth has a diagnosis of severe epilepsy, which is featured by often occuring seizures.
- Also, in those who have abnormalities in the brains, and which MRI does not show, the most often occuring cause is FCD.
- In kids who have undergone surgical treatment to combat the condition, FCD is the most often appearing trigger, and in adult individuals, it is the 1/3 most often.
- The MELD AI technology that classifies FCD can be used for any patient with a suspected symptom who is older than three years and who has previously done an MRI.
Scientists say this latest development will make it easier for doctors to diagnose abnormalities that are now slipping out of sight. In the future, it can allow a huge number of people suffering from epilepsy to recover from this illness.