We are pleased to announce a new publication from our research group!
Deep learning models for automated movement assessment often struggle when only limited training data is available — a common challenge in physiotherapy research, where collecting many high-quality recordings is costly and time-consuming. Our new paper presents a method that generates additional, realistic training examples by simulating movement variations through a musculoskeletal model. This approach respects the body’s natural range of motion and automatically labels the generated examples, helping AI systems generalize better to new patients and adapt more effectively with only a few examples per person.
📄 Now open access in the journal AI (MDPI):
https://www.mdpi.com/2673-2688/7/9/381
by Andreas Spilz, Heiko Oppel, and Michael Munz
#AI #Physiotherapy #Research #UlmUniversityOfAppliedSciences