A dual-modality workflow for quantifying microvascular structure in human temporal lobe epilepsy

Microvascular remodeling is implicated in the pathophysiology of drug-resistant temporal lobe epilepsy (TLE). However, characterizing these changes in adult human surgical tissue is methodologically limited by the accumulation of lipofuscin, an autofluorescent pigment that obscures microvascular structures and confounds standard automated quantification. To address this, we developed a dual-modality workflow optimized for archival human tissue obtained from drug-resistant TLE patients and one post-mortem control. This approach integrates hematoxylin and eosin (H&E) staining for precise anatomical subfield delineation with lectin-based fluorescence imaging and spectral unmixing, a computational technique that separates specific vascular signals from overlapping lipofuscin autofluorescence. Using QuPath, an open-source image analysis platform, we validated the accuracy of this automated workflow by comparing the microvascular area fraction (MAF) against manual ground-truth annotations in randomized regions of interest. The automated workflow demonstrated excellent concordance with manual assessment (Spearman's ρ = 0.997, p < 0.001). Application of this method to a cohort of TLE patients revealed substantial inter-patient heterogeneity. Exploratory analysis showed a positive but non-significant association between hippocampal MAF and preoperative seizure frequency (ρ = 0.68, p = 0.14), suggesting a possible relationship that requires confirmation in larger cohorts. Furthermore, the workflow successfully resolved subfield-specific vascular heterogeneity within sclerotic tissue that is typically obscured by whole-region average morphometry. This study establishes a robust, spectrally-unmixed morphometry pipeline that effectively resolves lipofuscin artifacts in adult human brain tissue, providing a necessary methodological foundation for future pathology-stratified investigations into cerebrovascular remodeling in epilepsy.