Sztuczna inteligencja w autoimmunologicznych chorobach neurologicznych: analiza bibliometryczna i mapa badań naukowych
Artificial Intelligence in Autoimmune Neurological Disorders: A Bibliometric Analysis and Research Landscape
W skrócie
[Preprint - wstępne wyniki] Naukowcy przeanalizowali ponad 1400 publikacji naukowych dotyczących zastosowania sztucznej inteligencji w leczeniu autoimmunologicznych chorób neurologicznych z lat 2015-2026. Okazało się, że większość badań dotyczy stwardnienia rozsianego i obrazów MRI, podczas gdy rzadsze choroby, badania EEG, markery biologiczne oraz łączne analizy danych pozostają znacznie niedoexploatowane. Liczba publikacji wzrasta szybko - z 10 prac w 2015 roku do prawie 300 w 2025 roku - co pokazuje rosnące zainteresowanie sztuczną inteligencją w neurologii.
Oryginalny abstract (angielski)
Abstract Artificial intelligence (AI) and machine learning are being applied rapidly across autoimmune neurological disorders, but the overall shape of this literature — which disorders, data modalities and methods dominate, and where evidence is absent — has not been quantified. Objective. To map the published research landscape of AI in autoimmune neurological disorders through a bibliometric analysis, and to identify the field’s concentrations and gaps. Methods. PubMed/MEDLINE was searched in November 2026 using a title/abstract strategy combining terms for autoimmune neurological disorders (multiple sclerosis [MS], NMOSD, MOGAD, autoimmune encephalitis, autoimmune epilepsy, paraneoplastic neurological syndromes) with artificial-intelligence and machine-learning terms, limited to 2015–2026. Records were analysed programmatically for annual output, leading journals, authorship, open-access status, and disorder-, modality- and method-level focus (title-level tagging). Results. 1,455 records were analysed. Annual output rose from 10 (2015) to 270 (2025), with 2026 already at 319 ahead-of-print records; output grew approximately 12-fold between 2015–2018 and 2022– 2025. The literature was dominated by multiple sclerosis (883 records, 60.7%), with all other disorders comparatively rare (NMOSD 45; autoimmune encephalitis 26; MOGAD 11; paraneoplastic 1). MRI was the dominant data modality (517 records, 35.5%), whereas EEG (9), genomics (25) and CSF biomarkers (83) were seldom addressed. Only 33 records (2.3%) described multimodal or data-fusion approaches. Deep learning (259) had overtaken classical machine learning (226). Work was distributed across 463 journals (led by Multiple Sclerosis and Related Disorders, n=84); 63.9% of records were open access. Conclusions. Research on AI in autoimmune neurology is expanding steeply but is heavily concentrated on multiple sclerosis and MRI. Rarer disorders, electrophysiological, fluid-biomarker and genomic data, and genuine multimodal integration remain markedly under-explored — defining clear priorities for future work.