An innovative system objectively assesses balance in blind people
A new study, led by the Universidad Politécnica de Madrid, proposes an objective assessment of balance in people with visual impairments using AI and inertial sensors.
09.10.2025
Millions of people with visual impairments face challenges trying to maintain balance, which affects their mobility and independence. In response, an international group of researchers has developed an innovative system that combines artificial intelligence and inertial sensors to objectively assess balance, especially in blind individuals. This is the first study which applies this technology in the context of the mini-BESTest (Mini Balance Evaluation Systems Test) clinical trial. The results obtained can help physical therapists obtain a more objective assessment of their patients' balance and develop telerehabilitation systems. The study, published in the journal PeerJ Computer Science, represents an important step toward more inclusive, technologically advanced, and precise rehabilitation.
The work was led by the Centro de Tecnología Biomédica (CTB) at the Universidad Politécnica de Madrid (UPM), specifically by the Bioinstrumentation and Nanomedicine Laboratory, with the participation of other institutions such as Universidad de La Laguna (ULL), Tenerife, la Université La Sagesse (ULS), Líbano; y the INDICATIC AIP association from Panamá.
The study was conducted at the CTB facilities and at the Madrid headquarters of the Organización Nacional de Ciegos de España (ONCE). It has been published in the journal PeerJ Computer Science and represents an important step toward more inclusive, technological, and precise rehabilitation.
The authors of the study compared the performance of blind and sighted individuals in a series of exercises designed to assess balance. To do so, they used a portable device with 12 motion sensors distributed on different parts of the study participants' bodies.
The most notable result was that the one-legged balance test, i.e., performing the "one leg" test, identified significant differences between the two groups. This task was much more difficult for blind individuals. Furthermore, it was observed that the physical therapists' assessments sometimes did not match the objective exercise duration data recorded by the sensors, suggesting some subjectivity in traditional clinical assessment.
Artificial intelligence models, especially those based on deep learning, were able to predict with 85.6% accuracy whether a person had good or poor balance in this test. “These results pave the way for the development of remote balance assessment tools, key to personalized and accessible telerehabilitation programs,” says José Javier Serrano Olmedo, a researcher at CTB-UPM. “The advances achieved with our work not only improve the way balance is measured in people with visual impairments but also lay the foundation for new forms of home rehabilitation that are more autonomous and based on objective data,” the researchers conclude.
The project code is available inGitHub (https://github.com/mjaenvargas/mini-BESTest_blind_noblind) and data in Zenodo (Jaén-Vargas, M., Pagán, J., Li, S., Kazemi, N., & Serrano-Olmedo, J. J. (2024). Automated balance assessment for blind and non-blind individuals using mini-BESTest and AI (v2.0.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.13842814).
Jaén-Vargas M, Pagán J, Li S, Trujillo-Guerrero MF, Kazemi N, Sansò A, Codina-Casals B, Abi Zeid Daou R, Serrano Olmedo JJ. 2025. AI-driven balance evaluation: a comparative study between blind and non-blind individuals using the mini-BESTest. PeerJ Computer Science 11:e2695 https://doi.org/10.7717/peerj-cs.2695
Jaén Vargas, M.Q. (2024) Contribution to methodologies for human activity recognition: Focusing on the visually impaired. Archivo Digital UPM. Available at: https://oa.upm.es/82209/ (Accessed: 03 July 2025).