Understanding, facilitating and predicting aphasia recovery after rehabilitation.

This is a narrative review, not a systematic review or meta-analysis, so the selection of studies may be selective. Many cited studies were small, including 3 patients for writing training and 7 participants in an ICCR study, so effects are not established. Several predictors relied on research-only measures such as fMRI connectivity, graph measures, lesion clustering, white matter hyperintensity ratings or app data, not routine clinic assessments. Treatment gains were often for trained items, and generalisation to discourse, functional communication or standardised cognitive outcomes was inconsistent. The review summarises heterogeneous populations, mostly chronic post-stroke aphasia and some acquired brain injury, so it may not apply to acute care or different therapies. Predictive models and BiLex recommendations were not yet validated for routine clinical use, and the randomised trial protocol was ongoing.

The supplied publication types indicate research support from NIH extramural and non-U.S. government sources. The article text does not state author conflicts or whether funders influenced design or reporting.

Do not read this review as showing that brain imaging or machine learning can already guide individual aphasia treatment. Most predictive measures came from small research studies, and some language therapy gains did not generalise beyond trained items.