CLSP-REQA: A Real-Time Quality-Aware Closed-Loop Seizure Prediction Framework with Mamba-BiLSTM and Confidence-Gated Intervention
TL;DR
This study is reported as a feasibility demonstration rather than evidence of prospective, real-world clinical readiness, and the comparison with domain adaptation should not be read as unconditional superiority.
Abstract
Reliable seizure prediction is a prerequisite for closed-loop neurostimulation therapy, yet existing methods rarely account for the variability in EEG signal quality encountered in real-world deployment, and the overwhelming majority adopt non-strict evaluation protocols that overestimate generalization performance. We propose CLSP-REQA (Closed-Loop Seizure Prediction with Real-time EEG Quality Assessment), a unified framework that embeds a lightweight signal quality estimator directly within the prediction pipeline. A Real-time EEG Quality Assessment (REQA) module runs in parallel with a Mamba-BiLSTM backbone, producing a scalar quality score q ∈ [0, 1] that modulates output confidence through a tiered non-linear fusion function (ECLO). We conducted dedicated experiments testing whether this learned score reflects genuine signal quality rather than a redundant proxy for the prediction task, and a sensitivity analysis of the fusion thresholds against standard calibration baselines. The framework was evaluated under strict cross-patient evaluation on the CHB-MIT Scalp EEG Database ( n = 23 subjects, 198 seizures) and the SIENA Scalp EEG Database ( n = 14 subjects, 47 seizures). On CHB-MIT, CLSP-REQA achieved an AUC-ROC of 0.7426 ± 0.0199, exceeding the unadapted cross-patient baseline of 0.69 reported by Jemal et al., using only 16 EEG channels compared to 23 in prior work, and without requiring any target-patient data or domain adaptation. On SIENA, CLSP-REQA achieved an AUC of 0.7012 ± 0.0249, exceeding the best domain-adapted cross-patient result of 0.61 on the same dataset under a strictly harder information constraint, since domain adaptation methods are given access to unlabelled target-patient data that CLSP-REQA does not use. The framework outputs a structured four-tuple 〈 p, q, c , Φ SHAP 〉 designed for compatibility with closed-loop neurostimulator interfaces. As both evaluation datasets are retrospective recordings from controlled hospital settings, we report this study as a feasibility demonstration rather than evidence of prospective, real-world clinical readiness, and the comparison with domain adaptation should not be read as unconditional superiority.
