Tele-EEG and AI-Enabled Digital Twins for Epilepsy: A Scoping Review

Epilepsy digital twins require patient-specific computational models that evolve with longitudinal observations and support prospective inference. We conducted a scoping review of complementary patient-specific modeling and longitudinal or remote EEG approaches, searching five databases with a publication cutoff of 26 August 2026. Additional retrieval on 11 October 2026 yielded 42 of the 54 previously unavailable reports. Following consistent eligibility reassessment, 68 reports were included, 22 were excluded after full-text assessment, and 12 remained unretrieved. Fourteen included reports were classified as C0, ten as C1, forty as C2, three as C3, and one as C4 using a review-specific functional framework. Three reports demonstrated repeated updating with prospective prediction or simulation, and one demonstrated experimental governed feedback. This combined count does not establish four longitudinally updated digital twins. No included report demonstrated the complete non-invasive longitudinal architecture of observation, explicit patient representation, repeated updating, uncertainty-aware prospective inference, and governed output. Mechanistic and dynamic network models contributed individualized representations, while remote EEG and adaptive computation contributed observation and prediction. Initial personalization was more common than repeated updating. The original single-reviewer workflow was supplemented by separate coauthor review of selection and functional classification. Incomplete retrieval, omitted Embase coverage, interpretive coding boundaries, and non-harmonized clinical descriptors still limit interpretation. Functional codes do not establish etiological applicability or treatment benefit. Progress requires population-specific, prospective validation of integrated architectures.