Preoperative Neuropsychology Subtypes Predict Neuropsychological Change after Temporal Lobe resection
TL;DR
Operating directly on routine preoperative evaluations without requiring additional testing, this computational framework disentangles pathological burden from network reserve, providing scalable, biologically interpretable biomarkers to guide personalized risk counselling and network-informed surgical planning.
Abstract
Cognitive outcomes following temporal lobe epilepsy surgery are highly heterogeneous and remain difficult to predict using traditional threshold-based neuropsychological classifications. Here, we implemented Subtype and Stage Inference (SuStaIn) on preoperative neuropsychology to model cognitive function as a continuous network-level process reflecting both pathological burden and compensatory reserve. We identified three distinct latent trajectories: Verbal, Naming, and Visual. The Verbal subtype reflected classic mesial temporal pathology, where postoperative decline aligned with functional adequacy of residual hippocampal tissue. Conversely, the Naming subtype represented a neocortical-predominant 'Temporal Plus' phenotype; despite lower rates of hippocampal sclerosis, these individuals showed severe postoperative verbal memory vulnerability due to un-reorganized frontotemporal language networks. The Visual trajectory demonstrated progressive visuospatial decline with distinct sex-specific reserve profiles and poorer visual recall outcomes. Crucially, these progression-based trajectories outperformed conventional static classifications in predicting 12-month postoperative outcomes on unseen test data. Operating directly on routine preoperative evaluations without requiring additional testing, this computational framework disentangles pathological burden from network reserve, providing scalable, biologically interpretable biomarkers to guide personalized risk counselling and network-informed surgical planning.