DisQ-HNet: A Disentangled Quantized Half-UNet for Interpretable Multimodal Image Synthesis Applications to Tau-PET Synthesis from T1 and FLAIR MRI

2026-02-25

AI Summary

This study addresses the challenge of limited accessibility and high cost of Tau-PET imaging in Alzheimer’s disease diagnosis by proposing DisQ-HNet, a framework that synthesizes interpretable Tau-PET images from multimodal T1 and FLAIR MRI. The method integrates a partial information decomposition (PID)-guided vector quantization encoder with a Half-UNet decoder, augmented by edge-aware pseudo skip connections to preserve fine anatomical details while disentangling and attributing modality-specific contributions. Experimental results demonstrate that DisQ-HNet outperforms existing baselines in reconstruction fidelity, more effectively retains disease-relevant Tau signals, and significantly enhances downstream tasks such as Braak staging, Tau localization, and classification. Furthermore, the framework provides clear, modality-specific interpretability, offering valuable insights into the contribution of each MRI sequence to the synthesized Tau-PET output.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Tau positron emission tomography (tau-PET) provides an in vivo marker of Alzheimer's disease pathology, but cost and limited availability motivate MRI-based alternatives. We introduce DisQ-HNet (DQH), a framework that synthesizes tau-PET from paired T1-weighted and FLAIR MRI while exposing how each modality contributes to the prediction. The method combines (i) a Partial Information Decomposition (PID)-guided, vector-quantized encoder that partitions latent information into redundant, unique, and complementary components, and (ii) a Half-UNet decoder that preserves anatomical detail using pseudo-skip connections conditioned on structural edge cues rather than direct encoder feature reuse. Across multiple baselines (VAE, VQ-VAE, and UNet), DisQ-HNet maintains reconstruction fidelity and better preserves disease-relevant signal for downstream AD tasks, including Braak staging, tau localization, and classification. PID-based Shapley analysis provides modality-specific attribution of synthesized uptake patterns.

Problem

• tau-PET synthesis
• multimodal MRI
• interpretable image synthesis
• Alzheimer's disease
• medical image translation

Innovation

• Disentangled representation
• Vector quantization
• Partial Information Decomposition
• Interpretable synthesis
• Half-UNet