Varatharajah’s Work Enables AI to Pinpoint Signs of Dementia in EEGs

Department of Computer Science & Engineering assistant professor Yoga Varatharajah is leveraging artificial intelligence (AI) and machine learning to analyze electroencephalogram (EEG) tests more efficiently. In a partnership with the Mayo Clinic, Varatharajah’s work has led neurologists to find early signs of dementia among the data analyzed in the EEGs.

EEG tests are typically done for people who have seizures or epilepsy. The tests examine the brain waves to see the types of seizures that are happening in a patient. In some cases, those patients might have dementia, but currently there is no way to identify signs of dementia using EEG. After analyzing EEGs from over 11,000 patients seen at the Mayo Clinic, Rochester, Varatharajah’s algorithm was able to spot distinct patterns in the patients with dementia that did not show up in other patients.

“We developed an unsupervised machine learning approach where we are mining the data without any specific goal,” said Varatharajah. “We took a large EEG database, and then we looked at the most common patterns that we see in this large population level data. We do find that there are certain patterns that are significantly different in dementia patients compared to cognitively normal individuals. We then used those patterns to build a supervised machine learning algorithm that classified these individuals into different diagnostic categories with very high accuracies.”

Varatharajah has been analyzing these EEGs and collaborating with the Mayo Clinic over the past eight years. Their previous work has used AI to extract insights that are more subtle than the results that can be observed by a trained clinician. This is a first of its kind study that looks at EEGs in the context of dementia based on the findings of Varatharjah’s quantitative methods. In the future, they hope to build on this process to develop more techniques that are able to capture the subtle patterns that may link to other diseases.

Generally, the diagnostic tests for dementia are based on imaging, like an MRI, and sometimes doctors opt for cerebrospinal fluid based measures which require a lumbar puncture. These tests are expensive and in the case of lumbar puncture, can be really painful. EEGs are a less expensive and more accessible way to diagnose dementia.

“Our goal here is not to replace standard dementia imaging tests with EEGs. We believe that, at medical facilities that lack expensive imaging equipment, it is plausible to use EEGs as a way to screen people that require more specialized care. We hope that, in the future, we can build on these findings to showcase a broader set of capabilities of these brain activity patterns, potentially in classifying dementia subtypes and more.”

Read more about the latest findings on the Mayo Clinic website.