Classification of Alzheimer's Patients through Ubiquitous Computing

Area of Science:

• Biomedical Engineering
• Data Science
• Neurology

Background:

• Parkinson's disease (PD) diagnosis and staging rely on clinical assessments.
• Objective, quantitative measures of motor function are needed for improved PD management.
• Smartphone sensors offer a feasible platform for collecting continuous movement data.

Purpose of the Study:

• To develop and validate a methodology for distinguishing movement patterns in Parkinson's disease patients across different disease stages.
• To classify new patients into their appropriate stage of Parkinson's disease using functional data analysis and artificial neural networks.

Main Methods:

• Utilized functional data analysis and artificial neural networks (ANNs) for pattern recognition.
• Collected three-dimensional movement data from Parkinson's disease patients using smartphone accelerometers during free movement.
• Applied the methodology to a novel, complex functional dataset with varying time domains, frequencies, and lengths.

Main Results:

• Achieved an 83% classification success rate in distinguishing between Parkinson's disease stages.
• Demonstrated the flexibility of the methodology across diverse and complex functional datasets.
• Successfully differentiated movement patterns indicative of different disease severities.

Conclusions:

• The proposed methodology shows significant potential for objective Parkinson's disease staging using smartphone data.
• Functional data analysis combined with ANNs provides a robust framework for analyzing complex movement patterns.
• This approach could facilitate more accurate and accessible Parkinson's disease assessment.