Longitudinal observation of psychophysiological data as a novel approach to personalised postural defect rehabilitation

• Published in: Scientific reports Vol. 15; no. 1; pp. 8382 - 15
• Main Authors: Romaniszyn-Kania, Patrycja, Pollak, Anita, Kania, Damian, Mitas, Andrzej W.
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 11.03.2025; Nature Publishing Group; Nature Portfolio
• Subjects: 639/166/985; 692/700/565/491; AdaBoost; Adult; Computer-assisted physiotherapy; Data analysis; Data processing; Emotion analysis; Female; Galvanic Skin Response; Humanities and Social Sciences; Humans; Longitudinal Studies; Machine learning; Male; Mathematical models; Middle Aged; multidisciplinary; Physiology; Posture; Posture - physiology; Precision Medicine - methods; Prediction models; Psychological assessment; Psychophysiological state analysis; Psychophysiology; Rehabilitation; Science; Science (multidisciplinary); Signal preprocessing; Statistical analysis; Therapeutic applications; Emotion analysis; Signal preprocessing; AdaBoost; Psychophysiological state analysis; Machine learning; Computer-assisted physiotherapy
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-025-92368-z

Postural defects are one of the main diseases reported to be at the top of the list of diseases of civilisation. The present study aimed to develop a novel approach to defining a set of measurable physiological biomarkers and psychological characteristics with identifiable information content and data analysis, enabling the determination of the adaptation period and conditioning the effectiveness of the treatment in personalised rehabilitation. During the rehabilitation, multimodal physiological signals (electrodermal activity, blood volume pulse) and psychological data (anxiety as a state and as a trait, temperament) were recorded on a group of 20 subjects over a period of three months (120 measurement sessions). Preprocessing of the physiological signals and psychological data was performed. A stepwise forward regression method was used to determine a set of successive statistically significant predictors of the model. For each group, a matrix of coefficients for fitting a linear regression of changes in the value of a given predictor in subsequent measurement was determined. Adaptive Boosting was chosen to develop a mathematical model of the patient. The analysis of the results of the psychological tests enabled the participants to be divided into five new, previously undefined subgroups, which were both labels for the classifier. Using the dimensionality reduction method, 8 significant, statistically important features were extracted. AdaBoost classifier allowed the creation of a prediction model for therapy parameters with 84% accuracy, and the Pseudo-Random Number Generator was used to check the validity of it. The AdaBoost algorithm was used again to check the dynamics of changes in regression coefficients for individual groups—a set of psychophysiological characteristics identified as the basis for personalised therapeutic interventions. Each individual requires time to adapt to a new situation, conditioned by their characteristics. An appropriate interdisciplinary approach to professional rehabilitation influences the therapeutic process’s quality, duration, and effectiveness. Physiological features determine the patient’s involvement in the rehabilitation process, allowing robust personalisation of therapy in a closed feedback loop. The fusion of psychophysiological data and multimodal measurements enables the development of a unique behavioral-physiological profile of the patient undergoing rehabilitation.