Real-Time Prediction of Correct Yoga Asanas in Healthy Individuals With Artificial Intelligence Techniques: A Systematic Review for Nursing.

Abstract

Aim: This study aims to systematically review the real-time prediction of yoga asanas using artificial intelligence (AI) techniques to improve the quality of life in healthy individuals.

Design: Systematic review.

Methods: A comprehensive literature review was conducted in English using the keywords 'yoga', 'asana', 'pose', 'posture', 'machine learning', 'deep learning' and 'prediction' in the Web of Science, Google Scholar, PubMed and Scopus databases. The objective was to identify all relevant studies on the topic. Two independent researchers screened the titles and abstracts of the retrieved publications, applying the JBI Critical Appraisal Checklist for Diagnostic Test Accuracy Studies for quality assessment. The initial search yielded 3250 studies (Google Scholar: 3190, PubMed: 19, Scopus: 27, Web of Science: 14). After applying inclusion criteria, 15 studies were included in the final systematic review.

Results: Among the included studies, nine employed deep learning (DL) models, three utilised machine learning (ML) and three applied a combination of both DL and ML techniques. The primary statistical evaluation method for real-time prediction was accuracy across all studies. The highest accuracy rates were observed in studies using DL models alone (min = 92.34%, max = 99.92%), followed by studies that combined DL and ML (min = 91.49%, max = 99.58%), and those using only ML (min = 90.9%, max = 98.51%). These findings indicate that integrating DL and ML models can enhance the accuracy of real-time yoga asana prediction.

Patient or public contribution: The findings advocate for the implementation of DL and ML models in clinical and community settings to improve the real-time and precise prediction of yoga asanas, a well-established evidence-based nursing intervention for healthy individuals.