AI model could help epilepsy patients free from seizures

A project led by Monash University The study, considered a world-first, shows that artificial intelligence (AI) models can potentially predict the best personalized anti-seizure medication for newly diagnosed epilepsy patients.

Predictive models, once fully developed, will save these patients from the uncertainty of not knowing when their lives will return to normal Sexuality by taking anti-seizure drugs, as well as the harmful side effects that may be associated with certain drugs.

Neuroscientist Professor Patrick Kwan Researchers from the Department of Neuroscience, Monash Central Clinical School are leading an international A collaboration that is “training” a deep learning predictive model (deep learning is a type of machine learning).

Their research was published in the influential JAMA Neurology .

Epilepsy affects 70 million people worldwide. Currently, choosing an anti-epileptic drug for a patient is a trial-and-error process, and clinicians cannot predict which drug a particular patient will respond to, Professor Guan said.

“If a patient does not respond to the first treatment, a significant number will respond to a second or third treatment There is a response, which means they might get rid of their seizures faster if the ‘right’ medication was chosen in the first place,” he said. “But if they’re taking the wrong medication, they can still have seizures, and they can also have side effects — they don’t benefit from the drug, they get hurt.”

These side effects range from allergies to psychiatric problems or, in the case of women of childbearing age, birth defects in babies. Some patients have drug-resistant epilepsy, which means that if predicted early, they can switch more quickly to other treatment options, including surgery, devices or diet, without wasting years with ineffective drugs.

The model used clinical information from 1,798 patients from five healthcare centers in Australia, Malaysia, China, and the United Kingdom. It was designed by Monash Medical AI led by Associate Professor Ge Zongyuan and trained using the Monash MASSIVE computing cluster.

“We are seeing how the latest deep learning models connect themselves from what is now computer-aided diagnosis to the field of therapy, which It’s really exciting,” said Associate Professor Ge.

The model’s accuracy in predicting the best drug was “moderate,” Kwan said. (It scores 0.65 on a statistical performance measure called AUROC, where 1.0 is the most accurate.) to train this base model. “ It has been improved both technically and by using more complex information. The augmented model will be Tested in the national multicenter randomized controlled PERSONAL trial (Personalized Choice of Drugs for Newly Diagnosed Adults with Epilepsy) to aid in epilepsy treatment selection.

Dr. Zhibin Chen, a Monash neuroscientist and biostatistician, played a key role in this study.

“This is considered a world-first model,” said Dr. Chen. “It ensures predictability in choosing the best treatment for newly diagnosed epilepsy patients.” It will open the door to personalized management of epilepsy. “

Dr. Harris Hakim, Alfred’s student and epilepsy researcher, is the lead author, While Dr. Kwan said students Wei Feng and Jiun Choong played key roles in developing the model.

Hope this Research could ultimately improve the management and treatment of epilepsy. It aims to predict response to treatment, not actual seizures.

Currently this model is suitable for adults with new-onset epilepsy who will be starting their first medication. It has not been tested in children.

This model will lay the groundwork for more models for epilepsy patients.