Machine Learning for Differentiating Essential Tremor: A Scoping Review

Background: Essential tremor (ET) is the most common movement disorder, affecting ~6% of adults over 65 1. Differentiating ET from other tremors remains clinically challenging due to overlapping features and variable presentation. Artificial intelligence (AI), particularly machine learning (ML), has emerged a potential tool to support neurologists by enhancing pattern recognition and complementing traditional assessments in complex cases. This is the first scoping review examining ML's potential role in distinguishing essential tremor from other tremor types. Methods: A systematic, scoping search was conducted using PubMed, Cochrane, and Scopus through April 2025, in accordance with PRISMA guidelines-ScR 2. Studies applying AI to distinguish ET from other tremors were included. Of 548 studies screened, 97 underwent full-text review, with data extracted from 46. Results: 46 included studies encompassed 6,051 patients, including 2,358 with ET. ML models utilized diverse inputs: accelerometers, gyroscopes, voice recordings, Archimedes spirals, EMG, and video. Common algorithms were included vector machines (18 articles), k-nearest neighbors (9 articles), and convolutional neural networks (8 articles). There was a high amount of heterogeneity in reporting data, severely limiting between study comparisons. Reported classification accuracies ranged from 60% to 100% (mean: 89%). However, heterogeneity in data types, methodologies, and reporting limited cross-study comparability. Conclusions: ML shows promise as a decision-support tool by recognizing tremor features that may complement, but not replace, expert clinical assessment, particularly in diagnostically ambiguous cases. To enable clinical adoption, future studies must address current heterogeneity, develop standardized datasets, implement automated preprocessing, and focus on clinically feasible data sources.