Deep multimodal learning for domain-level cognitive decline prediction in Alzheimer’s disease

Sep 28, 2026

Uploaded by Sarah

Alzheimer’s disease shows substantial variability in clinical progression, yet existing work only sparsely addresses prediction of cognitive decline. This study develops artificial intelligence models that learn neurocognitive trajectories from multimodal neuroimaging, using information from 653 ADNI participants. Models predict future clinical diagnoses and cognitive decline quantitatively and qualitatively. Results report strong diagnostic prediction and domain-wise deterioration detection, and show that modality-specific pre-training with late fusion improves accuracy.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Introduction

Methods

Results

Conclusions

TYPE Original Research PUBLISHED 17 December 2025 DOI 10. 3389/frai.2025.1731062 OPEN ACCESS EDITED BY Moolchand Sharma, Maharaja Agrasen Institute of Technology, India REVIEWED BY Vikash Yadav, Government Polytechnic Bighapur, Unnao, India Monica Bhutani, Bharati Vidyapeeth’s College of Engineering, India *CORRESPONDENCE Fernando García-Gutiérrez [fegarc05@ucm.es](fegarc05@ucm.es) RECEIVED 23 October 2025 REVISED 15 November 2025 ACCEPTED 24 November 2025 PUBLISHED 17 December 2025 CITATION García-Gutiérrez F, Matias-Guiu JA and Ayala JL (2025) Deep multimodal learning for domain-level cognitive decline prediction in Alzheimer’s disease. Front. Artif. Intell. 8:1731062 . doi: 10.3389/frai.2025.1731062 COPYRIGHT © 2025 García-Gutiérrez, Matias-Guiu and Ayala. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Deep multimodal learning for domain-level cognitive decline prediction in Alzheimer’s disease Fernando García-Gutiérrez1*, Jordi A. Matias-Guiu2 and José L. Ayala1 for the Alzheimer’s Disease Neuroimaging Initiative (ADNI) 1 Department of Computer Architecture and Automation, Universidad Complutense de Madrid, Madrid, Spain, 2 Department of Neurology, Hospital Clínico San Carlos, Instituto de Investigación Sanitaria San Carlos (IdISSC), Madrid, Spain Introduction: Alzheimer’s disease (AD) is characterized by signiﬁcant variability in clinical progression; however, few studies have focused on developing models to predict cognitive decline. Anticipating these trajectories is essential for patient management, care planning, and developing new treatments. This study explores the potential of artiﬁcial intelligence (AI) techniques to model neurocognitive trajectories from multimodal neuroimaging data and further investigates different data representation frameworks. Methods: Using information from 653 participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), we developed models to predict future clinical diagnoses and cognitive decline, both quantitatively (rate of decline) and qualitatively (presence or absence of decline) . Input features included structural T1-weighted magnetic resonance imaging (MRI), [ 18 F]-ﬂuorodeoxyglucose positron emission tomography (FDG-PET), [ 18 F]-ﬂorbetapir PET (AV45-PET), neuropsychological assessments, and demographic variables. Several information representation strategies were explored, including tabular data models, convolutional neural networks (CNNs), and graph neural networks (GNNs) . Furthermore, to maximize the use of all available information, we proposed a modeling framework that performed modality-speciﬁc pre-training to learn feature embeddings, which were then integrated through a late-fusion layer to produce a uniﬁed representation for downstream prediction. Results: The modeling strategies demonstrated good predictive performance for future clinical diagnoses, consistent with previous studies (F1 = 0. 779) . Quantitative models explained approximately 29.4%–36.0% of the variance in cognitive decline. In the qualitative analysis, the models achieved AUC values above 0.83 when predicting cognitive deterioration in the memory, language, and executive function domains. Architecturally, CNN- and GNN-based models yielded the best performance, and the proposed pre-training strategy consistently improved predictive accuracy. Conclusions: This study demonstrates that AI techniques can capture patterns of cognitive decline by exploiting multimodal neuroimaging data. These ﬁndings contribute to the development of more precis

FAQs

What problem does the study address in Alzheimer’s disease?

It targets prediction of cognitive decline and future clinical diagnoses despite high variability in disease progression.

Which data modalities and inputs are used for the models?

The models use structural T1-weighted MRI, FDG-PET, AV45-PET, neuropsychological assessments, and demographic variables.

What modeling framework is proposed to improve performance?

The study proposes modality-specific pre-training to learn feature embeddings, then integrates them via a late-fusion layer for downstream prediction.

Sep 28, 2026

Uploaded by Sarah