Validation and Development of a Posture Quantification and Classification Technique Using MediaPipe Pose®

ITURRIETA, JUAN; PÉREZ, ELISA; TETTAMANTI, FERNANDO; LÓPEZ, NATALIA

IFMBE PROCEEDINGS

Springer Nature

Año: 2024 vol. 105 p. 280 - 288

1680-0737

Currently, according to the WHO, neurological disorders leave a high rate of motor sequelae in many individuals worldwide. For public health, this represents a great cost that many developing countries cannot absorb in their health system, or are overwhelmed to provide a prompt response to these demands. The importance of proposing therapies to support adequate rehabilitation makes us think of methods that can be more effective and low-cost in neurorehabilitation. Therefore, being able to classify positions objectively in a therapy environment is of utmost importance, helping subjectivity of the therapist, or of the therapy itself, and reducing time and costs. In this paper we present a development on quantification and classification of upper limb postures, using a webcam-type sensor and MediaPipe®. The algorithm developed to estimate angular excursion using the Bland-Altman method was validated and it can be concluded that it is interchangeable with the goniometer.