Artificial Intelligence techniques applied to rehabilitation of patients with musculoskeletal and cognitive disorders
| dc.contributor.advisor | García-Zapirain,Begoña | es_ES |
| dc.contributor.advisor | Méndez Zorrilla, Amaia | es_ES |
| dc.contributor.author | Shapoval, Serhii | es_ES |
| dc.date.accessioned | 2025-02-28T09:26:28Z | |
| dc.date.available | 2025-02-28T09:26:28Z | |
| dc.date.issued | 2024-12-19 | |
| dc.description.abstract | This thesis explores the potential of applying and implementing deep learning methods in the field of medical physical and cognitive rehabilitation of users with age or cognitive disabilities. In this context, two case studies have been conducted: one investigating the prospects and possibilities of applying the Serious Games family of integrated deep learning algorithms to the rehabilitation and support of elderly people, and the other focusing on the support and training of categories of people with cognitive disabilities. The first case study analyzes the physical condition of users who fall into the age category of 60 years or more, as well as progress as a result of prolonged implementation of prescribed rehabilitation measures. In this study, a group of 15 people were asked to perform a standardized set of physical exercises focused on working different muscle clusters and joints. There were two sets of exercises: the standard set including ordinary exercises, and the second set via the Serious Game application, in which the same exercises were interpreted as game tasks. The peculiarity of the proposed application was that it integrated a modified algorithm based on a trained Neural Network, which allows to register the user's movements via web-camera. As a result of performing a course of exercises in two interpreters during two days, positive results were achieved in three main research parameters: time and accuracy of exercise performance, as well as the overall activity index. In the course of analyzing the results, it was found that the use of assistant applications has a positive impact on users, and in contrast to conventional exercises, there is an increase in results. Thus, the accuracy of exercise results when using the assistant system increases from 54.7% to 85.4%, while when performing the same exercises in the standard interpretation, on the contrary, it decreases from 44.3% to 36.2%. More dramatic changes are observed in time indices. When compared to the reference time required to complete a particular exercise (Exercise 4 as an example), the difference in time results when using the application decreases from 45.3% to 14.63%, while when performing standard exercises it increases from 55.7% to 63.8%. In terms of overall gains, user results improve by 18-40% on average, depending on the user and the exercise in question. In the second case study the impact of the introduction and use of a course of daily use of software-assistants in the processes of support and training of users with peculiarities of cognitive state is investigated. In this case, testing was conducted on 4 groups of users, totaling 54 people. The groups were gathered according to 2 parameters: presence of cognitive peculiarities, level of cognitive disability. An additional parameter was age, which ranged from 20 to 58 years old. The level of cognitive disability also varies from 5% to 97%. During the testing, the participants were required to use the developed assistant application, which had 2 levels of 5 tasks each. As a result, only one group completed the two levels, at 100% and 92% respectively. The rest of the groups completed only Level 1 tasks. Also, all results were analyzed based on 6 main evaluation parameters: age, gender, level of disability, number of correct answers, time to complete the task, and number of clicks made during the task execution. According to the correlation analysis, the highest values were found in the relationship between the parameters of correct answers and level of disability, correct answers and time, time and age. Further analysis of clustering and evaluation using various methods of statistical analysis, including Pearson correlation coefficient evaluation method, ordinary least squares (OLS), random forest method, support vector machine model showed that all the parameters presented above are interrelated and are in mutual dependence, as evidenced by the index of 0.93 out of 1.00. In addition, the subsequent additional user survey showed that this interpretation of the tasks is more acceptable and interesting. | es_ES |
| dc.identifier.uri | http://hdl.handle.net/20.500.14454/2397 | |
| dc.language.iso | eng | es_ES |
| dc.publisher | Universidad de Deusto | es_ES |
| dc.subject | Ciencias Tecnológicas | es_ES |
| dc.subject | Tecnología de los ordenadores | es_ES |
| dc.subject | Dispositivos de transmisión de datos | es_ES |
| dc.subject | Matemáticas | es_ES |
| dc.subject | Ciencia de los ordenadores | es_ES |
| dc.subject | Inteligencia artificial | es_ES |
| dc.title | Artificial Intelligence techniques applied to rehabilitation of patients with musculoskeletal and cognitive disorders | es_ES |
| dc.type | doctoral thesis | es_ES |
Archivos
Bloque original
1 - 1 de 1