Neuropathology of focal epilepsy: the promise of artificial intelligence and digital Neuropathology 3.0.

Focal lesions of the human neocortex often cause drug-resistant epilepsy, yet ​surgical resection of the epileptogenic region has been proven as a successful strategy to control seizures in a carefully selected patient cohort. Continuous efforts to study neurosurgically resected brain samples at the microscopic level, i.e., Neuropathology 1.0, unravelled a comprehensive description of the spectrum of underlying aetiologies, e.g., hippocampal sclerosis, congenital brain tumours or cortical malformations as the three most common aetiologies representing almost 80% of the entire lesional landscape. Human brain tissue was also instrumental to discover underlying molecular pathways and common somatic variants, e.g., MTOR, DEPDC5, SLC35A2, BRAF or PTPN11, that helped us to define specific phenotype-genotype associations, thereby promoting novel targets for medical treatment, i.e., Neuropathology 2.0. The increasing gap in accessing necessary resources to perform such studies around the world could be bridged, however, when introducing artificial intelligence (AI)-based algorithms to classify epileptogenic brain lesions on digital slide scans obtained from routine haematoxylin and eosin-stained, formalin-fixed paraffin-embedded tissue sections. This may also provide an advanced prediction of the lesion's phenotype-genotype association in the near future. Thus, digital Neuropathology 3.0 may be the promising next level of laboratory advancement in the realm of neuropathology in focal epilepsy.