Detection of Postural Control in Young and Elderly Adults Using Deep and Machine Learning Methods with Joint-Node...

Area of Science:

• Biomechanics
• Gerontology
• Computer Science

Background:

• Postural control significantly declines with aging, necessitating reliable assessment tools.
• Current methods for evaluating postural control may lack efficiency or accuracy, especially in older populations.

Purpose of the Study:

• To develop and validate an efficient and accurate method for detecting postural control ability in elderly adults.
• To compare the performance of various deep and machine learning techniques for postural control classification.

Main Methods:

• Recruited 35 elderly adults and 20 young adults for standing tasks.
• Utilized Kinect device (30 Hz) to capture 15 joint node coordinates.
• Developed joint-node plots (JNPs) and applied 15 deep and machine learning methods for classification.

Main Results:

• The JNP method combined with machine learning achieved high performance metrics (accuracy, sensitivity, specificity, PPV, NPV, kappa > 0.9) in validation.
• The highest values for all assessed metrics exceeded 0.9, indicating robust detection capabilities.
• The JNP approach demonstrated strong performance in distinguishing postural control abilities between young and elderly participants.

Conclusions:

• The presented method using joint-node plots (JNPs) offers a highly accurate and efficient approach for assessing postural control.
• This technique shows significant promise for evaluating balance in both young and elderly populations.
• The integration of JNPs with machine learning provides a powerful tool for gerontological research and clinical applications.