Truace
Subscribe

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

Abstract Objective Artificial intelligence (AI) classifies routine electroencephalograms (EEGs) with expert‐level accuracy, but interpretation of long‐term EEG monitoring (LTM) remains reliant on time‐intensive visual review. We evaluated whether autoSCORE, a clinically approved implementation of SCORE‐AI integrated into an EEG reader, achieves expert‐level diagnostic accuracy for LTM interpretation and whether expert‐supervised use reduces review time versus conventional analysis. Methods A fixed‐and‐frozen AI…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1818· cite & verify
Artificial intelligence‐assisted categorization of interictal abnormalities in adult long‐term electroencephalographic monitoring: Comparable diagnostic accuracy with marked efficiency gain
Hospital Universitari Doctor Peset, València 09 by 19Tarrestnom65. CC BY-SA 4.0 · https://creativecommons.org/licenses/by-sa/4.0

In brief

Researchers tested autoSCORE, an approved version of SCORE-AI built into an EEG reader, on 78 adult long-term EEG recordings from one center. Two panels of three experts each read the recordings, one reviewing the AI's detections and the other reading the full recordings in the usual way.

The result suggests that supervised AI could ease the heavy visual-review burden of long-term monitoring while keeping or improving diagnostic quality. The evidence comes from a small single-center dataset, and the AI working alone performed worse than the hybrid approach, so human oversight appears to be essential.

Main points

  1. The test set had 78 adult recordings, 52 normal and 26 abnormal, with epilepsy monitoring unit discharge diagnosis as the reference standard.
  2. Interrater reliability was .81 for the hybrid approach versus .43 for conventional reading.
  3. Hybrid sensitivity was .88 versus .69 for conventional interpretation, with an adjusted p of .006.
  4. Hybrid accuracy of .75 beat fully automated analysis at .55, but its edge over conventional accuracy (.68) was nonsignificant after correction.

The gain

Expert review of autoSCORE detections shortened long-term EEG interpretation from a median of 52.4 to 2.2 minutes, improved interrater agreement, and raised sensitivity over conventional reading.

The rundown

The study used a fixed-and-frozen AI model, so the algorithm was not adjusted during testing. It was benchmarked against two expert panels, one using the hybrid approach and one the conventional approach, and time to diagnostic decision was compared across the fully automated, hybrid and conventional methods.

Median review time fell from 52.4 minutes (IQR 33.2–62.9) to 2.2 minutes (IQR .5–4.3), and mean time fell from 50.8 to 2.9 minutes. Statistical comparisons of diagnostic performance were adjusted for 63 tests using the Holm–Bonferroni procedure.

The authors conclude that expert oversight of AI categorization of interictal epileptiform and nonepileptiform abnormalities improves reliability and efficiency, raises sensitivity over conventional reading, and raises accuracy over fully automated analysis.

Sources

  1. Peer-reviewedEpilepsia2026-10-17

The debate