Age Group Discrimination via Free Handwriting Indicators

Area of Science:

• Gerontology
• Biomedical Engineering
• Machine Learning

Background:

• Ageing is linked to cognitive and functional decline, impacting daily living.
• Early detection of unhealthy ageing is crucial but challenging due to similarities with normal ageing.
• Chronic diseases can accelerate age-related decline.

Purpose of the Study:

• To develop an early screening method for healthy ageing using handwriting analysis.
• To assess handwriting performance in different age groups (40-59, 60-69, 70+).
• To differentiate between normal ageing and potentially unhealthy decline.

Main Methods:

• Utilized an instrumented ink pen to collect raw handwriting data from 60 healthy subjects.
• Extracted fourteen indicators related to gesture and tremor from handwriting.
• Employed machine learning algorithms for binary classification tasks across age groups.

Main Results:

• Achieved high accuracy (97.5%), F1 score (97.44%), and ROC-AUC (95%) in distinguishing between early-stage ageing and elderly subjects.
• Identified age-dependent sensitivity in handwriting and tremor indicators.
• Shapley value analysis confirmed the significance of specific indicators.

Conclusions:

• The proposed handwriting analysis method shows promise for early detection of abnormal ageing signs.
• This non-invasive, remote monitoring approach can improve care for older adults.
• The technology supports unsupervised home monitoring for proactive health management.