INESC TEC researchers develop chat to support physiotherapy processes

It is called Physio: an innovative chat system that provides physical rehabilitation information to support patients and physiotherapists. The solution was developed by INESC TEC researchers and received several awards.

Using language models in areas like physiotherapy may seem abstract and contradictory; but a group of INESC TEC researchers proved otherwise, developing a chat system that combines both areas, using generative models to support physical rehabilitation.

Bringing “physiotherapy to everyone” is the major advantage of Physio, a chat tool developed by INESC TEC researchers Hugo Sousa, Ruben Almeida, Filipe Cunha, Nuno Guimarães, Ricardo Campos and Alípio Jorge.

Created in just one week, “the project was developed within the scope of a competition promoted by Sword AI – whose challenge was to develop an innovative product using natural language models”, explained Hugo Sousa. The proposal ended up winning the third place in the competition.

According to the researcher, Physio is nothing more than a software that “stands out for the ability to generate personalised replies to users, based on reliable sources, and referencing them in said replies”. In addition, “it is equipped to recommend specific exercises and, if necessary, medication, making it a valuable resource for both patients and healthcare professionals”, he guaranteed.

Thanks to the innovative and relevant nature of the project – integrating language models into real-world applications -, the group of researchers won the Best Demo Paper Award at the 46^th edition of the European Conference on Information Retrieval (ECIR’24), organised by the University of Glasgow; this event acknowledges the best scientific papers with demonstration of applications and software for innovative scientific ideas.

Regarding the next steps, Hugo Sousa does not hide the desire to improve the project with the specialised support of physiotherapists, to increase the reliability of the model and further safeguard the well-being of the patients. “It would be interesting to expand the agent’s knowledge base, integrating more curated sources. In addition, it would be interesting to perform clinical validation to ensure that the replies provided by Physio comply with the best practices”, he concluded.

The researchers mentioned in this news piece are associated with INESC TEC, UP-FCUP and UBI.