Exploring Risk Factor Interactions across the Development Stages of Dementia using an Explainable Machine Learning Model

Aug 5, 2026

Upload by Maeve

Early prediction of dementia, a long-term progressive neurodegenerative disorder, remains challenging. Recent AI-based diagnostic tools often lack clinical interpretability because they produce oversimplified outputs and behave as black boxes. This study assigns clinically meaningful categories to a longitudinal cohort dataset and trains an interpretable random forest model. It further applies an advanced explanation framework to analyze predictions, identify impacts of key risk factors, and compare interaction patterns across development stages.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Introduction

Methodology

Data Description and Preprocessing

Dan so, S a mu el O an d Luo, Ze qi (2 0 2 4) Exploring Risk Factor Inter a ction s a cro s s th e Developm ent St ag e s of Dem entia u sing an Explain able M achin e Learning Mod el. In: 2 0 2 4 2 9 th Intern ation al Confer enc e on Autom ation an d Computing . IEEE, pp . 1-6 . ISBN 9 7 9-8-3 5 0 3-6 0 8 8-2 Downloa d e d from : [http :// sur e . sun d erl an d. ac . uk/id/ e print / 1 7 9 9 0 /](http :// sur e . sun d erl an d. ac . uk/id/ e print / 1 7 9 9 0 /) U s a g e g uid eli n e s Ple a s e r efer to th e u s a g e guid eline s at [http :// sur e . sun d erl an d . ac. uk/ policies. html](http :// sur e . sun d erl an d . ac. uk/ policies. html) or altern atively cont act [sur e @ sun d erlan d. ac . uk](sur e @ sun d erlan d. ac . uk). Exploring Risk Factor Interactions across the Development Stages of Dementia using an Explainable Machine Learning Model Zeqi Luo School of Computing Science University of Glasgow Glasgow, United Kingdom [z.luo.3@research.gla.ac.uk](z.luo.3@research.gla.ac.uk) Samuel O. Danso School of Computer Science University of Sunderland Sunderland, United Kingdom [sam.danso@sunderland.ac.uk](sam.danso@sunderland.ac.uk) Abstract—Early prediction of dementia, a long-term progressive disease, has always been a challenge. In recent years, advances in artificial intelligence have led to new computeraided diagnostic tools. However, these methods often offer limited interpretability due to their simplistic binary outputs and blackbox algorithms, restricting their use. In this work, we addressed aforementioned shortcomings by assigning clinically meaningful categories to a longitudinal cohort dataset and using an interpretable random forest algorithm to train the prediction model. Our results show that the model predicts various categories effectively. We further applied an advanced machine learning explanation framework to analyse the predictions, revealing the impact of some key risk factors on the prediction and varying interaction patterns between these factors when predicting different development stages of dementia. Index Terms—brain health, dementia, risk factors, machine learning I. INTRODUCTION Dementia is a complex neurodegenerative disorder known tobe associated with a variety of symptoms including cognitive decline, which results in loss of independence and function [1] . Alzheimer’s disease (AD) is the most common type of dementia and it is estimated to account for over 70 percent of dementia cases [2] . While AD has no cure, there is adequate evidence that points to the fact that the pathophysiological process a result of accumulation of amyloid in the brain begins up to 20 years before the clinical manifestation of the disease. This evidence has led to categorisation of the stages of dementia based clinical manifestation to facilities early detection. This categorisation is important is it provides a framework for early detection. The framework enables clear characterisation of the various stages of the disease process for early detection and intervention. While clinical trials involving the stages of mild to moderate dementia have failed to demonstrate clinical usefulness, it is believed that early detection of dementia risk from the preclinical stage will have clinical benefit. Computational approaches such as Machine Learning have demonstrated to have the potential to accurately predict the dementia risk and assist in the diagnosis of AD [3] . However, majority of these approaches have focused on the later stages of the disease, missing the critical window. A recent research by Danso et al. [4] focused on early detection up to 14 years of dementia onset. The authors defined dementia risk based on family history and apolipoprotein E type 4 allele status. In this work, we develop an explainable early detection machine learning model based on the Amyloid/Tau/Neurodegeneration (ATN) classification framework [5] . The ATN approach offers the possibility to

FAQs

Why is early dementia prediction difficult, according to the document?

Early prediction is challenging because dementia is a complex neurodegenerative disorder and many AI approaches provide limited interpretability. Black-box or overly simplistic outputs restrict clinical usefulness.

What modeling approach is used to improve interpretability?

The work assigns clinically meaningful categories to a longitudinal cohort dataset and trains an interpretable random forest algorithm. It then uses a machine learning explanation framework to analyze prediction drivers.

How are risk factors studied across dementia development stages?

The study applies explanation methods to reveal which key risk factors influence predictions. It also examines how interaction patterns between risk factors vary across different development stages of dementia.

Aug 5, 2026

Upload by Maeve