Reliability of Bonini's semiological classification in frontal lobe seizures: A video-EEG based study

Frontal lobe seizures have heterogeneous semiology, making electroclinical localization challenging. Bonini et al. proposed a semiological classification based on rostrocaudal frontal lobe organization. We evaluated its interobserver and intraobserver reliability and explored anatomical and surgical correlations.

We retrospectively analyzed long-term video-EEG recordings of 33 patients with suspected frontal lobe epilepsy. Two independent observers classified seizures into four Bonini groups and repeated the classification after 12 weeks. Semiology was interpreted using standardized ILAE terminology where applicable, while group allocation was based on the descriptors of the original Bonini classification. Reliability was assessed using Cohen's kappa, percentage agreement, and an interobserver confusion matrix. Neuroimaging findings were mapped to Brodmann areas and correlated with Bonini group allocation as an exploratory analysis.

Intraobserver reliability was excellent, with kappa values of .869 and .914. Interobserver agreement was substantial, with kappa values of .6535 and .6988 across the two trials. Group 2 seizures were most frequent, whereas Group 4 was rare. The confusion matrix showed predominant diagonal agreement, with disagreements mainly between adjacent groups. Among 13 patients with clearly localizable focal lesions, semiology correlated with anatomical localization in nine. Nine patients underwent surgery; concordance was observed in seven, and all achieved Engel Class I outcomes.

Bonini's classification shows excellent intraobserver reliability and substantial interobserver agreement in routine video-EEG practice. Anatomical validation remains limited by the absence of systematic SEEG confirmation, but the observed imaging and surgical correlations support its clinical applicability as a reproducible framework for presurgical hypothesis generation.

This is the first study to validate the reproducibility of Bonini's classification in routine video-EEG practice; however, network-based anatomical interpretation should remain cautious in the absence of systematic SEEG validation.