Pattern detection from seating pressure distribution during wheelchair motion using deep embedded clustering.

Area of Science:

• Biomedical Engineering
• Data Science
• Clinical Nursing

Background:

• Minimizing pressure ulcers in high-risk individuals, like wheelchair users, requires identifying typical pressure distribution patterns.
• Detecting these patterns is difficult due to the variability of pressure distribution during movement compared to static postures.

Purpose of the Study:

• To develop and evaluate a method for detecting typical pressure distribution patterns associated with pressure ulcers.
• To assess the effectiveness of deep embedded clustering for this pattern identification task.

Main Methods:

• Utilized deep embedded clustering, a technique that extracts features using an auto-encoder and optimizes data points into clusters.
• Compared deep embedded clustering with traditional methods (k-means, PCA+k-means) on a dataset of 26,944 pressure distribution images labeled by nursing experts.