Scoping review of ALS databases: Opportunity for multimodal machine learning

Document Type : Review Article

Authors

1 Sean M Healey & AMG Center for ALS, Harvard Medical School, Massachusetts General Hospital, Boston, MA, USA

2 Department of Neurosurgery, School of Medicine, Alborz University of Medical Sciences, Karaj, Iran Clinical Research Development Unit, Shahid Madani Hospital, Alborz University of Medical Sciences, Karaj, Iran Neuroscience Research Center, Shahid Beheshti University of Medical Sciences, Tehran. Iran

Abstract
Background: Amyotrophic lateral sclerosis (ALS) is a progressive, incurable neurodegenerative disease. ALS databases play a critical role in supporting research, decision-making, and the development of generative machine learning (ML) applications. This scoping review aims to explore genetic/imaging/facial expression/voice databases that can inform ALS disease research, supporting future multimodal ML applications.
Methods: A literature search was conducted across PubMed, Scopus, Web of Science, Cochrane Library, and grey literature, covering publications up to September 2025. The Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) guidelines were followed. Two reviewers independently screened titles, abstracts, and full texts. Databases were included if they provided ALS-specific data on genomics, medical imaging, voice, or facial expression. Extracted information included data type, accessibility, demographic details, key features, limitations, and database link. This review identifies gaps in the current database landscape. The Newcastle-Ottawa Scale (NOS) was used for evaluating study quality.
Results: Twenty-four ALS databases were identified, accounting for twenty studies of moderate quality. Most resources were single-modality, with only a few [e.g., Sporadic ALS Australia Systems Genomics Consortium (SALSA-SGC), Answer ALS, Canadian ALS Neuroimaging Consortium (CALSNIC)] offering multidimensional data. Common limitations included restricted demographic diversity, small sample sizes in digital cohorts, limited longitudinal follow-up, and insufficient cross-modal integration.
Data gaps limit the development of accurate ML models in ALS.
Conclusion: This scoping review provides a comprehensive understanding of ALS databases for future multimodal deep learning research. Research gaps were highlighted, and recommendations for future directions were stated.

Graphical Abstract

Scoping review of ALS databases: Opportunity for multimodal machine learning

Keywords

Subjects


Articles in Press, Accepted Manuscript
Available Online from 16 August 2026