A new Australian study conducted by Monash University, in collaboration with Alfred Health and the Royal Melbourne Hospital, reveals how machine learning technology could be used to automate the diagnosis of epilepsy.
In this study, researchers at Monash University applied more than 400 electroencephalogram (EEG) recordings of patients with and without epilepsy from Alfred Health and Royal Melbourne Hospital to a model machine learning. Learning the model with the different data sets allowed it to automatically detect signs of epilepsy – or abnormal activities known as “spikes” in EEG recordings.
“The objective of the first step is to evaluate existing models involved in the detection of abnormal electrical recordings among neurons in the brain, called epileptiform activity. These anomalies are often sharp peaks that stand out from the rhythmic patterns of a patient’s EEG, ”says Levin Kuhlmann, senior lecturer at Monash University, in the computer science department of the faculty of data and science. IA.
Doug Nhu, research associate of the project and doctoral student of the faculty, specifies that applying machine learning to the process could free up the time of health professionals, because the current process of diagnosing epilepsy is often long. “Being able to apply a machine learning model to various datasets demonstrates our ability to create an algorithm that is more reliable, more adaptive and smarter than existing models, which makes our model more useful when applied in real scenarios, like the diagnosis of patients in a clinic, ”he explains.
In addition to diagnosing epilepsy patients, the technology has potential as a training tool for graduate neurologists, who can use it as a basis for comparison with records of epilepsy patients, the university adds.
“Our plans for this research will be to continue to improve current models and to continue training them against additional data sets from other hospitals,” said Patrick Kwan, of the Department of Neuroscience, Faculty of Medicine. Monash University. “We aim to develop an accurate algorithm that will be reliable in multiple hospital environments and usable in the early stages of epilepsy diagnosis, based on routine EEG recordings and sleep deprived EEG recordings. “
According to Levin Kuhlmann, the next stage of the project will see the machine learning model focus on new seizure detection and prediction methods.