Scoping review of machine learning models for post-stroke aphasia diagnosis and outcome prediction

ObjectiveTo systematically review the literature on the application of machine learning models in post-stroke aphasia, and to provide a reference for the construction and clinical application of related models.MethodsBased on scoping review methodology, we searched Web of Science, PubMed, Cochrane Library, Embase, CINAHL, CNKI, VIP database, Wanfang database, and China Biology Medicine. The search time limit was from the database’s establishment to November 20, 2025, and the retrieved literature was screened, summarized, extracted, and analyzed.ResultsA total of 19 articles were included. The analysis results showed that the machine learning algorithms used in post-stroke aphasia models were mainly supervised methods, including random forests, neural networks, and support vector machines. The data sources of the model were diverse. The indicators included in the model covered multimodal data. The functions of the model include diagnosis and classification of aphasia patients, assessment and prediction of the severity of aphasia patients, prediction of the language function and rehabilitation outcome of patients, monitoring and evaluation of symptoms, etc.ConclusionMachine learning models have high applicability and broad scope in post-stroke aphasia. Future research still requires multi-center, multi-modal data and external validation to enhance its robustness and clinical feasibility.