Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Class Balancing Diversity Multimodal Ensemble for Alzheimer's Disease Diagnosis and Early Detection

Arianna Francesconi, Lazzaro di Biase, Donato Cappetta, Fabio Rebecchi, Paolo Soda, Rosa Sicilia, Valerio Guarrasi

TL;DR

This work tackles Alzheimer’s disease diagnosis and early detection by fusing multimodal, tabular data from the ADNI database while addressing severe class imbalance. It introduces IMBALMED, an ensemble framework that trains classifiers on diverse class-balanced subsets across four modalities and applies two-level fusion (unimodal then multimodal) to yield robust predictions. Across binary and ternary diagnostic tasks and 12-, 24-, 36-, and 48-month early detection, IMBALMED outperforms an unbalanced baseline and nine state-of-the-art imbalance methods, with strongest gains in the clinically crucial 48-month horizon. The method demonstrates stable performance and highlights the value of balancing diversity and multimodal integration for AD risk stratification, though future work should validate on external datasets and explore modality-specific ablations.

Abstract

Alzheimer's disease (AD) poses significant global health challenges due to its increasing prevalence and associated societal costs. Early detection and diagnosis of AD are critical for delaying progression and improving patient outcomes. Traditional diagnostic methods and single-modality data often fall short in identifying early-stage AD and distinguishing it from Mild Cognitive Impairment (MCI). This study addresses these challenges by introducing a novel approach: multImodal enseMble via class BALancing diversity for iMbalancEd Data (IMBALMED). IMBALMED integrates multimodal data from the Alzheimer's Disease Neuroimaging Initiative database, including clinical assessments, neuroimaging phenotypes, biospecimen and subject characteristics data. It employs an ensemble of model classifiers, each trained with different class balancing techniques, to overcome class imbalance and enhance model accuracy. We evaluate IMBALMED on two diagnostic tasks (binary and ternary classification) and four binary early detection tasks (at 12, 24, 36, and 48 months), comparing its performance with state-of-the-art algorithms and an unbalanced dataset method. IMBALMED demonstrates superior diagnostic accuracy and predictive performance in both binary and ternary classification tasks, significantly improving early detection of MCI at 48-month time point. The method shows improved classification performance and robustness, offering a promising solution for early detection and management of AD.

Class Balancing Diversity Multimodal Ensemble for Alzheimer's Disease Diagnosis and Early Detection

TL;DR

This work tackles Alzheimer’s disease diagnosis and early detection by fusing multimodal, tabular data from the ADNI database while addressing severe class imbalance. It introduces IMBALMED, an ensemble framework that trains classifiers on diverse class-balanced subsets across four modalities and applies two-level fusion (unimodal then multimodal) to yield robust predictions. Across binary and ternary diagnostic tasks and 12-, 24-, 36-, and 48-month early detection, IMBALMED outperforms an unbalanced baseline and nine state-of-the-art imbalance methods, with strongest gains in the clinically crucial 48-month horizon. The method demonstrates stable performance and highlights the value of balancing diversity and multimodal integration for AD risk stratification, though future work should validate on external datasets and explore modality-specific ablations.

Abstract

Alzheimer's disease (AD) poses significant global health challenges due to its increasing prevalence and associated societal costs. Early detection and diagnosis of AD are critical for delaying progression and improving patient outcomes. Traditional diagnostic methods and single-modality data often fall short in identifying early-stage AD and distinguishing it from Mild Cognitive Impairment (MCI). This study addresses these challenges by introducing a novel approach: multImodal enseMble via class BALancing diversity for iMbalancEd Data (IMBALMED). IMBALMED integrates multimodal data from the Alzheimer's Disease Neuroimaging Initiative database, including clinical assessments, neuroimaging phenotypes, biospecimen and subject characteristics data. It employs an ensemble of model classifiers, each trained with different class balancing techniques, to overcome class imbalance and enhance model accuracy. We evaluate IMBALMED on two diagnostic tasks (binary and ternary classification) and four binary early detection tasks (at 12, 24, 36, and 48 months), comparing its performance with state-of-the-art algorithms and an unbalanced dataset method. IMBALMED demonstrates superior diagnostic accuracy and predictive performance in both binary and ternary classification tasks, significantly improving early detection of MCI at 48-month time point. The method shows improved classification performance and robustness, offering a promising solution for early detection and management of AD.

Paper Structure

This paper contains 17 sections, 4 equations, 3 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Materials
  4. Dataset and Pre-Processing
  5. Patient Selection
  6. Methods
  7. Data Balancing
  8. Data Fusion Levels
  9. Experimental setup
  10. Training and Evaluation
  11. Competitors and Statistical Analysis
  12. Results and Discussions
  13. Multimodal Fusion
  14. Competitors and Statistical Analysis
  15. Conclusions
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Overview of selected data modalities from the ADNI dataset. Each modality is depicted in a labeled box, accompanied by an icon representing the data type and specific tests included in the modality. Arrows indicate the flow of data from the ADNI database to its integration into our final dataset. For notation, see Section \ref{['sec:balancing']}.
  • Figure 2: Schematic representation of the proposed method, illustrating two main steps: training, which comprises data balancing and classifier training, and testing, which involves unimodal and multimodal fusion to determine class membership probabilities.
  • Figure 3: IMBALMED vs. Unbalanced Dataset Approach for various tasks. Panels (a) and (b) represent the binary and ternary diagnostic tasks, respectively. Panels (c) to (f) represent the 12-, 24-, 36-, and 48-month early detection tasks, respectively. Lines connect classifiers' average performance.