FAV graduate won 2nd place and Audience Award with rehabilitation system at the IT SPY competition

Tomáš Kment developed a device to improve exercising for patients with upper-limb mobility issues. His thesis earned him 2nd place and the Audience Award in the 16th IT SPY competition, standing out among 1,500 entries.

It recognizes in real-time when a person with an upper limb mobility impairment is about to move and helps them rehabilitate. Such a system was developed by Tomáš Kment, now a graduate of the Faculty of Applied Sciences at the University of West Bohemia in Pilsen (FAV), who chose the topic for his thesis. The system processes EEG signals, which it uses to communicate with a robotic arm, both in virtual reality and in a computer application. Customized neurorehabilitation software can help in the treatment of patients with nervous system injuries, Parkinson's disease, and other neurological disorders.

What makes it unique is that Tomáš Kment used affordable equipment, such as a commonly available EEG cap and free third-party software. He assembled and tested a complete end-to-end system. "An end-to-end model means that it can convert input to output without manually defining individual steps. The input is the measured data on the brain activity of the patient undergoing rehabilitation, and the output is the analysis of the data obtained and the appropriate connection of the robotic arm so that the rehabilitation is effective and fun for the patient," explained Tomáš.

The test results confirmed that it is difficult to generalize brain activity because it always depends on the patient's experience and skills. At the same time, Tomáš managed to identify things that are often overlooked in the literature. "For the correct interpretation of the outputs, it is necessary to clearly separate the patient data used to train the model from the data of other patients used to confirm it. In some cases, all patient data is combined into one large training set, which leads to overly optimistic results. However, these do not reflect reality," explained Tomáš, pointing out a possible pitfall.

Supervisor Roman Mouček emphasized that Tomáš's approach was not only technical, but also methodologically well thought out: "What I appreciate about Tomáš's work is that he did not succumb to the desire to just 'show good numbers'. On the contrary, he very openly identified the limitations of current approaches and proposed realistic steps to move this area forward. It is precisely this kind of critical thinking that is key to transferring results into practice."

The FAV graduate also focused on problematic moments in data collection, such as blinking, chewing, or turning the head, which cause distortion of brain waves. Thanks to comprehensive research, Tomáš's work not only provides software and data but also specific methodological recommendations on how to test this area more realistically and put it into practice with the appropriate knowledge of the context.

For ethical and legislative reasons, it was not possible to test the system directly on patients with physical disabilities. The current data is therefore based on simulated conditions and testing on healthy volunteers. "If clinical studies in the future confirm that the model and the entire system can reliably improve the performance of rehabilitation, it would pave the way for their use in hospitals and rehabilitation centers," added Roman Mouček.

IT SPY is an elite competition for the best thesis in the field of computer science and information technology. Every year, 20 Czech and Slovak universities participate in the competition, submitting up to 1,500 defended theses. FAV graduates regularly occupy the top positions in this prestigious competition, confirming that FAV belongs to the elite club of the best IT faculties in Czechia and Slovakia.