Desenvolvimento e comparação entre diferentes implementações de lógica fuzzy para aptidão física em reabilitação de joelho

Abstract

Artificial Intelligence (AI) since its beginning had as a strategy to mimic human cognition. In the area of physical rehabilitation, the studies include AI to process, estimate and classify the level of physical activity to improve the professional–patient relationship. The purpose of this work was to develop and compare Sugeno (FLS) and Mamdani (FLM) Fuzzy Logic (FL) implementations to assist the physiotherapist’s decision to let the patient returns to activities with more data and in a safer way. The implemented systems are composed of fuzzy rules (if – then) and four inputs of range of motion, extension and flexion; pain intensity; and muscle strength; to generate an output on the physical capability of the knee. The qualitative requirements of the systems took into account processing time, precision and reliability of the responses. In the comparison between FLM and FLS, the Sugeno method was more reliable regarding the level of membership function, but both systems agreed on the values reported in six hypothetical clinical cases and the resulting capability concepts. To the assessment of these systems, three physical therapists responded to the same six clinical cases and their responses were compared with the conceptual outputs of the FLM and FLS systems, here considered as a template to test the output. The level of agreement was high in two of the six cases and low in the other four, as an innovative method of assessment in the rehabilitation area, it is believed that with more data being part of the systems and with professionals having access to it, the agreement may be higher. Several clinical tests can be added as inputs to assist in the professional decision-making process, with new studies to validate, streamline, and evolve the treatment provided.