26 Αυγ 2026
New iPROLEPSIS publication presents machine learning approach to personalise game difficulty based on patient condition
A new iPROLEPSIS publication presents a system that automatically adjusts game difficulty to match each patient's changing condition during therapeutic gameplay.
The study proposes a Dynamic Difficulty Adjustment (DDA) system using machine learning – specifically Reinforcement Learning and Multi-Armed Bandits –to personalise how two serious games challenge players: Mr. Blue Sky, targeting wrist mobility, and The Kite, focusing on breathing and relaxation.
Data were collected from 32 participants playing these games, tracking gameplay performance, physiological signals (heart rate and stress), and self-reported relaxation levels. Machine learning algorithms learned which difficulty progression works best based on how patient condition changes during play.
Results showed that the best-performing strategies began gameplay at medium difficulty, then adjusted downward. In a feasibility pilot with six psoriatic arthritis patients, this automatic adjustment produced coherent therapeutic progressions while accounting for the fluctuating nature of patient symptoms.
The system demonstrates how machine learning can support personalised rehabilitation, continuously adapting game challenges to patient needs rather than using fixed difficulty levels.
Read the full publication: https://ieeexplore.ieee.org/document/11663158
