Applications of machine learning and natural language processing to neurocognitive outcomes in posttreatment cancer...

This scoping review explores how machine learning (ML) and natural language processing (NLP) are used to detect, characterize, and predict neurocognitive symptoms in cancer survivors across age groups. The review had two goals: (1) to compare ML and NLP applications in understanding cancer-related cognitive impairment (CRCI) among age-stratified survivors and (2) to identify research gaps that could inform future survivorship care.

Following PRISMA-ScR guidelines, a comprehensive literature search was conducted across PubMed, Scopus, Web of Science, IEEE Xplore, and Google Scholar from 2014 to 2025. Studies were included if they used ML or NLP to assess neurocognitive outcomes in posttreatment cancer survivors. Studies without defined ML/NLP methods, a survivorship focus, or peer review were excluded.

The final review included 27 studies with 3584 participants. Most studies used supervised ML models such as random forest and support vector machines. Key applications included predicting patient-reported outcomes and identifying biomarkers via neuroimaging. Most studies focused on adult survivors, with limited research in older adult (n = 4), AYA (n = 1), and pediatric (n = 3) populations specifically, despite their high risk for long-term CRCI.

ML and NLP show promise for CRCI detection. Future research should prioritize developing age-specific ML/NLP models for underrepresented populations, particularly older adults, AYA, and pediatric survivors, while establishing standardized validation frameworks. Additionally, interdisciplinary collaboration and integration into clinical workflows will be essential for effective implementation.