Study: AI Assistance Improves Clinicians’ EEG Interpretation Accuracy

AI Assistance Improves Clinicians’ EEG Interpretation Accuracy in Study

A recent cross-over study conducted by Barnett and colleagues has shown promising results in improving clinicians’ electroencephalogram (EEG) interpretation accuracy with the assistance of artificial intelligence (AI). The study, published in NEJM AI, compared the performance of clinicians with and without AI assistance and found a significant increase in accuracy when using AI.

Seizures are serious medical events that can have severe consequences, and accurate interpretation of EEGs is crucial in diagnosing and treating them. However, clinician availability and subjectivity can hinder accurate diagnoses. To address this issue, Barnett and his team developed a deep-learning algorithm called ProtoPMed-EEG, trained on data from over 2700 hospitalized patients.

In the study, eight clinicians without specialized EEG or machine-learning expertise were randomly assigned to two groups. Each group was given ProtoPMed-EEG at different stages, two weeks apart, and asked to interpret 100 EEG samples. The results showed that the mean diagnostic accuracy with AI assistance was 71%, compared to 47% without AI (p < 0.05). Additionally, the mean inter-rater reliability improved with AI assistance. Despite the longer time needed to diagnose with AI, clinicians reported that their diagnostic ability improved with the use of the AI model. In a post-study survey, 7 out of 8 clinicians recommended the AI model for educating future medical professionals. Overall, this study demonstrates the potential of AI to assist clinicians in making more accurate diagnoses and improving inter-rater reliability in EEG interpretation. The success of ProtoPMed-EEG in this study suggests that AI could play a significant role in diagnostic assistance and clinical education in the future.