An Unsupervised Analysis of an Alzheimer’s Disease Patient Population Using Subspace Search and Hierarchical Density-Based Clustering

Syndromes that affect the mental health of individuals, such as Dementia and Mild Cognitive Impairment, are characterized by degradation of cognitive functions. The consequences affect not only the individual’s life but also family and caregivers. Alzheimer’s disease is the most common cause of these disorders. Cognitive assessments (screening tests) are used worldwide, particularly in poor regions of the world, to diagnose such diseases. However, those assessments are subject of individual analysis by clinicians, and so can be challenging to diagnose given the subtle aspects and symptoms of the illness. In this work, we present an unsupervised analysis using subspace search to find subspaces of highly correlated features and a hierarchical density-based algorithm to find patterns in a dataset of an Alzheimer’s disease patient population. We show that the subspace search can help discover correlation relationships between features and that subspaces can be used to find meaningful clusters of individuals. We also show that structurally well-formed clusters may not lead to good models when the features are not relevant to the domain.