Artificial Intelligence-enabled Adaptation in Rehabilitation Games: A Systematic Review

In rehabilitation games, recognising a patient’s state and adapting the game are not the same thing. This review asked how participant and task data move through AI methods into patient-facing adaptive changes, and whether evaluation designs can isolate the value of that adaptation. That distinction shaped what evidence we needed to extract and compare.

The team identified 29,777 records across six databases/information sources and included 32 studies from 2016 through May 2026. As third author, I participated across the PRISMA workflow, completed my assigned screening, and produced structured classification tables for synthesis. The search corpus is the team’s total, not an individual screening count.

My work involved data curation, validation, and original-draft preparation, as well as conceptualisation and investigation. I helped distinguish systems that only classify or predict a state from those whose AI outputs change the intervention. Following search logic, eligibility, extraction, classification, and synthesis end to end gave me a practical foundation for later review-design work.

Research capability: turn heterogeneous technical studies into a structured, traceable evidence synthesis.

AI-enabled adaptation takes many forms, but controlled clinical evidence isolating its unique effect remains limited. The review is complete and the manuscript has been submitted; the supplied record does not identify a confirmed venue or decision. I present it as a submitted research contribution, not an accepted publication or a claim of clinical effectiveness.