A study team examined the reliability of human assessments versus an automated algorithm in evaluating the quality of intracranial electroencephalography (iEEG) data.
The research, featured in the Journal of Neural Engineering, was led by the University of Minnesota Medical School, focusing on how human experts compare to an automated algorithm when assessing iEEG data quality. The aim of this study is to improve seizure detection and localization, thereby benefiting patients with epilepsy.
iEEG is a technique that records brain activity through electrodes placed on or within the brain. This precise data is essential for diagnosing and managing epilepsy, particularly for identifying the exact source of seizures to enable effective treatment.
In this investigation, the team gathered 16 specialists, including EEG technologists and neurologists trained through fellowships, to evaluate 1,440 iEEG channels, categorizing them as either “good” or “bad.” A “good” rating indicated the channel was accurately recording brain activity, while “bad” indicated it wasn’t. Their assessments were compared against each other and against the Automated Bad Channel Detection (ABCD) algorithm, created by the Herman Darrow Human Neuroscience Lab at the University of Minnesota.
The ABCD algorithm showed a higher accuracy rate of 95.2% and performed better overall than human evaluators, especially in detecting channels with significant high-frequency noise.
“Our results reveal potential biases and shortcomings in evaluations conducted by humans. The effectiveness of the ABCD algorithm hints at a future where automated tools can assist healthcare providers in enhancing the precision and efficiency of seizure detection, thereby improving patient care,” remarked Alexander Herman, MD, PhD, who is an assistant professor at the U of M Medical School and attending psychiatrist with M Health Fairview.
This study emphasizes the promise of automated technologies to bolster the reliability and efficiency of interpreting iEEG data, which is critical for accurately locating seizures and improving patient outcomes.
“This research illustrates that automated algorithms have the capability to surpass human experts in detecting poor-quality EEG channels. By lightening the load and reducing variance in assessments, we can shift our focus towards clinical decision-making and patient care,” stated David Darrow, MD, MPH, an assistant professor at the U of M Medical School and a neurosurgeon with M Health Fairview.
Further research should work on enhancing these automated approaches and investigate their use in real-time clinical environments.
The study received funding from the Institute for Translational Neuroscience and MnDRIVE Brain Conditions.
