Temporal Pattern Mining for Analysis of Longitudinal Clinical Data: Identifying Risk Factors for Alzheimer's Disease

TL;DR

This paper introduces a novel framework combining temporal abstraction, pattern mining, and survival analysis to identify predictive risk factors for Alzheimer's disease from complex longitudinal clinical data.

Contribution

It presents the first application of survival analysis on temporal AD data using pattern mining, enhancing interpretability and predictive accuracy.

Findings

Achieved a Concordance index of up to 0.8 in predicting AD.

Discovered meaningful temporal patterns associated with AD risk.

Provided a visualization tool for interpretability of patterns.

Abstract

A novel framework is proposed for handling the complex task of modelling and analysis of longitudinal, multivariate, heterogeneous clinical data. This method uses temporal abstraction to convert the data into a more appropriate form for modelling, temporal pattern mining, to discover patterns in the complex, longitudinal data and machine learning models of survival analysis to select the discovered patterns. The method is applied to a real-world study of Alzheimer's disease (AD), a progressive neurodegenerative disease that has no cure. The patterns discovered were predictive of AD in survival analysis models with a Concordance index of up to 0.8. This is the first work that performs survival analysis of AD data using temporal data collections for AD. A visualisation module also provides a clear picture of the discovered patterns for ease of interpretability.