Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Radiogenomic Bipartite Graph Representation Learning for Alzheimer's Disease Detection

Aditya Raj, Golrokh Mirzaei

TL;DR

This paper tackles Alzheimer's disease detection by integrating structural MRI with gene-expression data from three driver genes APOE, PSEN1, and PSEN2. It introduces a radiogenomic bipartite graph learning framework composed of a 3D denoising autoencoder for MRI feature extraction, a bipartite graph that fuses imaging and genomics, and a heterogeneous graph neural network that learns a dynamic adjacency and edge weights for classification. The approach achieves state-of-the-art like performance on AD vs CN, AD vs MCI, and CN vs MCI tasks (e.g., up to $Acc \approx 92\%$, $F1 \approx 93\%$ for AD vs CN) and provides interpretable gene contributions via averaged edge-weights. Overall, the method demonstrates that radiogenomic fusion with dynamic graph learning can improve diagnostic accuracy and reveal biologically meaningful gene importance, with potential applicability to other diseases using radiogenomic data.

Abstract

Imaging and genomic data offer distinct and rich features, and their integration can unveil new insights into the complex landscape of diseases. In this study, we present a novel approach utilizing radiogenomic data including structural MRI images and gene expression data, for Alzheimer's disease detection. Our framework introduces a novel heterogeneous bipartite graph representation learning featuring two distinct node types: genes and images. The network can effectively classify Alzheimer's disease (AD) into three distinct stages:AD, Mild Cognitive Impairment (MCI), and Cognitive Normal (CN) classes, utilizing a small dataset. Additionally, it identified which genes play a significant role in each of these classification groups. We evaluate the performance of our approach using metrics including classification accuracy, recall, precision, and F1 score. The proposed technique holds potential for extending to radiogenomic-based classification to other diseases.

Radiogenomic Bipartite Graph Representation Learning for Alzheimer's Disease Detection

TL;DR

This paper tackles Alzheimer's disease detection by integrating structural MRI with gene-expression data from three driver genes APOE, PSEN1, and PSEN2. It introduces a radiogenomic bipartite graph learning framework composed of a 3D denoising autoencoder for MRI feature extraction, a bipartite graph that fuses imaging and genomics, and a heterogeneous graph neural network that learns a dynamic adjacency and edge weights for classification. The approach achieves state-of-the-art like performance on AD vs CN, AD vs MCI, and CN vs MCI tasks (e.g., up to , for AD vs CN) and provides interpretable gene contributions via averaged edge-weights. Overall, the method demonstrates that radiogenomic fusion with dynamic graph learning can improve diagnostic accuracy and reveal biologically meaningful gene importance, with potential applicability to other diseases using radiogenomic data.

Abstract

Imaging and genomic data offer distinct and rich features, and their integration can unveil new insights into the complex landscape of diseases. In this study, we present a novel approach utilizing radiogenomic data including structural MRI images and gene expression data, for Alzheimer's disease detection. Our framework introduces a novel heterogeneous bipartite graph representation learning featuring two distinct node types: genes and images. The network can effectively classify Alzheimer's disease (AD) into three distinct stages:AD, Mild Cognitive Impairment (MCI), and Cognitive Normal (CN) classes, utilizing a small dataset. Additionally, it identified which genes play a significant role in each of these classification groups. We evaluate the performance of our approach using metrics including classification accuracy, recall, precision, and F1 score. The proposed technique holds potential for extending to radiogenomic-based classification to other diseases.
Paper Structure (7 sections, 6 equations, 3 figures, 1 table)

This paper contains 7 sections, 6 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. 3D Denoising Autoencoder for Feature Extraction
  4. Radiogenomic Bipartite Graph Construction
  5. Bipartite Graph Neural Network
  6. Experiments and Results
  7. Conclusion

Figures (3)

  • Figure 1: The proposed framework of bipartite graph representation learning with imaging genomics data.
  • Figure 2: The bipartite graph construction for imaging-genomics fusion in AD detection. (A) graph containing all samples in the study. (B) Subgraph representing one sample.
  • Figure 3: The heterogeneous bipartite GNN model with dynamic edge weight learning.