Detecting Cognitive Impairment Status Using Keystroke Patterns and Physical Activity Data among the Older Adults: A...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Gerontology

Background:

• Cognitive impairment significantly impacts global healthcare and communities, especially among older adults.
• Aging often leads to declines in cognition and mental retention, making early detection vital to prevent permanent mental damage.

Purpose of the Study:

• To develop a machine learning model for detecting and differentiating cognitive impairment severity (severe, moderate, mild, normal).
• To analyze neurophysical and physical data for improved cognitive impairment prediction.

Main Methods:

• Extracted neurophysical data via keystroke dynamics and physical data via smartwatches.
• Employed the Gradient Boosting Machine (GBM) ensemble learning algorithm for classification.
• Utilized Pearson's correlation and wrapper feature selection for optimal feature identification.

Main Results:

• The proposed GBM model achieved an accuracy exceeding 94% in classifying cognitive severity.
• Successfully integrated neurophysical and physical data for enhanced prediction capabilities.

Conclusions:

• The study demonstrates the efficacy of a machine learning approach combining neurophysical and physical data for accurate cognitive impairment detection.
• This research offers a novel dimension to the state-of-the-art in predicting cognitive decline.