Predicting Alzheimer’s Disease Diagnosis Risk Over Time with Survival Machine Learning on the ADNI Cohort

Aug 4, 2026

The rise of Alzheimer’s Disease worldwide drives demand for efficient tools to forecast cognitive decline and dementia risk. This paper investigates Survival Machine Learning as a modeling approach that predicts both deterioration and the likely time until deterioration. Using the ADNI cohort, the study shows strong predictive performance and reliable calibration, supporting its use for clinical investigation. The results indicate that time-dependent risk estimation can add actionable information beyond conventional classification models when assessing Alzheimer’s Disease progression.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages14

Reading Time35 min

AccessFree to download

Introduction

Survival analysis and censored data

Motivation for time-dependent risk prediction

Study aims and dataset context

Paper structure

Musto, Henry; Stamate, Daniel; Pu, Ida and Stahl, Daniel. 2023. 'Predicting Alzheimer's Disease Diagnosis Risk Over Time with Survival Machine Learning on the ADNI Cohort'. In: Computational Collective Intelligence. ICCCI 2023 .. Budapest, Hungary 27–29 September 2023 . [Conference or Workshop Item] [https://research.gold.ac.uk/id/eprint/35866/](https://research.gold.ac.uk/id/eprint/35866/) The version presented here may differ from the published, performed or presented work. Please go to the persistent GRO record above for more information. If you believe that any material held in the repository infringes copyright law, please contact the Repository Team at Goldsmiths, University of London via the following email address: [gro@gold.ac.uk](gro@gold.ac.uk). The item will be removed from the repository while any claim is being investigated. For more information, please contact the GRO team: [gro@gold.ac.uk](gro@gold.ac.uk) Predicting Alzheimer’s Disease Diagnosis Risk over Time with Survival Machine Learning on the ADNI Co hort Henry Musto1 , Daniel Stamate1, 2 , Ida Pu 1, Daniel Stahl 3 1 Data Science & Soft Computing Lab, Computing Department, Goldsmiths College, University of London, United Kingdom 2 Division of Population Health, Health Services Research & Primary Care, School of Health Sciences, University of Manchester, United Kingdom 3 Institute of Psychiatry Psychology and Neuroscience, Biostatistics and Health Informatics De partment, King’s College London, London, UK Abstract. The rise of Alzheimer’s Disease worldwide has prompted a search for efficient tools which can be used to predict deterioration in cognitive decline leading to dementia. In this paper, we explore the potential of Survival Machine Learning as such a tool for building models capable of predicting not only deterioration but also the likely time to deterioration. We demonstrate good predictive ability and calibration, lending support to its use in clinical investigation and prediction of Alzheimer’s Disease risk. Keywords: Survival Machine Learning, ADNI, Clinical Prediction Modelling. 1 Introduction One of the most pressing challenges for governments and healthcare systems is the rising number of people with dementia. More than 55 million people live with dementia worldwide, and there are nearly 10 million new cases yearly, with 60-70% of all dementias being of Alzheimer’s Disease type (AD) [1] . Recently, attention has turned to machine learning (ML) as a tool for improving the predictive ability of clinical models concerning AD and addressing clinical challenges more widely. However, of the hundreds of clinical ML models that appear in scientific publications every years, few have thus far been successfully embedded into existing clinical practice [2] . Oneof the reasons why machine learning has not seen success in clinical practice is that it is unable to predict the risk of the disease occurring. This means that, although clinicians may be afforded a prediction of who is likely to develop a disease, they are notable to quantify the risk of disease development as time passes [3] . In the context of predicting the progression of AD in particular, many studies that use ML methods employ a classification approach, whereby the outcome to be predicted is either a binomial or multinomial outcome [4][5] . The datasets are often derived from longitudinal studies, whereby clinical marker data is collected from participants over months and years [6] . Thus, such data have a temporal element inherent to 2 the methodology employed in the collection process. However, standard classification ML cannot consider the predictive power of time in conjunction with other predictors. With this in mind, a newly emerging field of exploration seeks to build on traditional time-dependent statistical models, such as survival analysis, to develop machine learning models which can predict the time-dependent risk of developing AD and go

FAQs

What problem does the paper address in Alzheimer’s Disease prediction?

It targets the need to predict deterioration in cognitive decline and to estimate the risk of Alzheimer’s Disease over time rather than only classifying outcomes.

Why is survival analysis important for this task?

Survival analysis models time to an event and handles censored data, where the outcome has not occurred during the study for some participants.

How does Survival Machine Learning differ from standard classification approaches?

Standard classification models treat outcomes as binary or multinomial and do not incorporate time-dependent predictive power with other predictors, whereas Survival Machine Learning integrates time-to-event risk estimation.

Aug 4, 2026

Upload by Ava Thompson