AI-assisted review speeds interpretation of long-term EEG recordings

In a single-center diagnostic accuracy study of 78 adult long-term EEG recordings, expert-supervised autoSCORE review was compared with conventional expert interpretation and fully automated analysis. The hybrid approach substantially shortened review time and improved sensitivity versus conventional interpretation, although its accuracy advantage did not remain significant after multiple-comparison correction.

Hybrid review reduced median review time by 95.8%.
Sensitivity was higher than with conventional interpretation.
Accuracy advantage over conventional review was not significant.
Study included 78 adult recordings at one center.

Researchers evaluated a fixed, frozen version of autoSCORE, an AI system integrated into an EEG reader, using 78 adult long-term EEG monitoring recordings: 52 classified as normal and 26 as abnormal. The reference standard was the epilepsy monitoring unit discharge diagnosis. Three-reader panels assessed either AI-detected findings in a hybrid workflow or the full recordings conventionally; fully automated results were also assessed.

Median review time was 2.2 minutes with the hybrid approach, compared with 52.4 minutes for conventional review—a 95.8% median reduction. Interrater reliability was higher with hybrid review (Gwet’s AC1 0.81 vs 0.43; p<.001). Hybrid interpretation had greater sensitivity than conventional interpretation (0.88 vs 0.69; adjusted p=.006), and greater accuracy than fully automated analysis (0.75 vs 0.55; adjusted p<.001). Its accuracy compared with conventional interpretation was numerically higher (0.75 vs 0.68), but the difference was not significant after correction.

The findings support further evaluation of expert-supervised AI for adult long-term EEG review, but do not establish performance across other centers or populations. The study was single-center and included 78 recordings; the abstract does not provide additional details on generalizability or external validation.

In practice

For adult long-term EEG monitoring, expert-supervised review of autoSCORE detections was associated with markedly shorter review time, higher interrater reliability, and higher sensitivity than conventional interpretation in this single-center dataset. The hybrid approach also outperformed fully automated analysis in diagnostic accuracy. These results suggest potential workflow value, but the study included only 78 recordings from one center, and hybrid accuracy was not significantly higher than conventional accuracy after correction. The findings do not establish generalizability or support changing practice on their own.

AI-generated clinical interpretation · Not editor-reviewed

AI summary · Not yet editor-reviewed

Is this summary clinically accurate?

Help fellow clinicians: your rating sends inaccurate summaries straight to our editors.

Source

Epilepsia: Artificial intelligence‐assisted categorization of interictal abnormalities in adult long‐term electroencephalographic monitoring: Comparable diagnostic accuracy with marked efficiency gain ↗

This is an automated AI-condensed summary that has not yet been reviewed by an editor. Always consult the full item at the original source.