Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cas-DiffCom: Cascaded diffusion model for infant longitudinal super-resolution 3D medical image completion

Lianghu Guo, Tianli Tao, Xinyi Cai, Zihao Zhu, Jiawei Huang, Lixuan Zhu, Zhuoyang Gu, Haifeng Tang, Rui Zhou, Siyan Han, Yan Liang, Qing Yang, Dinggang Shen, Han Zhang

TL;DR

The paper tackles missing data in longitudinal infant brain MRI, which hinders reliable developmental analysis and atlas construction. It introduces Cas-DiffCom, a two-stage cascaded diffusion framework with multimodal guidance (image and age) and an asynchronous multimodal denoising block to ensure temporal consistency, followed by a conditional 3D super-resolution refinement. Key contributions include the AsMM block for disentangling multimodal guidance, dense longitudinal completion with high fidelity on the Baby Connectome Project dataset, and demonstrated usefulness for downstream tasks like tissue segmentation and developmental trajectory estimation. The results show substantial fidelity improvements (PSNR 24.15, SSIM 0.81) and trajectories that align with real data, signaling practical impact for infant neurodevelopment studies.

Abstract

Early infancy is a rapid and dynamic neurodevelopmental period for behavior and neurocognition. Longitudinal magnetic resonance imaging (MRI) is an effective tool to investigate such a crucial stage by capturing the developmental trajectories of the brain structures. However, longitudinal MRI acquisition always meets a serious data-missing problem due to participant dropout and failed scans, making longitudinal infant brain atlas construction and developmental trajectory delineation quite challenging. Thanks to the development of an AI-based generative model, neuroimage completion has become a powerful technique to retain as much available data as possible. However, current image completion methods usually suffer from inconsistency within each individual subject in the time dimension, compromising the overall quality. To solve this problem, our paper proposed a two-stage cascaded diffusion model, Cas-DiffCom, for dense and longitudinal 3D infant brain MRI completion and super-resolution. We applied our proposed method to the Baby Connectome Project (BCP) dataset. The experiment results validate that Cas-DiffCom achieves both individual consistency and high fidelity in longitudinal infant brain image completion. We further applied the generated infant brain images to two downstream tasks, brain tissue segmentation and developmental trajectory delineation, to declare its task-oriented potential in the neuroscience field.

Cas-DiffCom: Cascaded diffusion model for infant longitudinal super-resolution 3D medical image completion

TL;DR

The paper tackles missing data in longitudinal infant brain MRI, which hinders reliable developmental analysis and atlas construction. It introduces Cas-DiffCom, a two-stage cascaded diffusion framework with multimodal guidance (image and age) and an asynchronous multimodal denoising block to ensure temporal consistency, followed by a conditional 3D super-resolution refinement. Key contributions include the AsMM block for disentangling multimodal guidance, dense longitudinal completion with high fidelity on the Baby Connectome Project dataset, and demonstrated usefulness for downstream tasks like tissue segmentation and developmental trajectory estimation. The results show substantial fidelity improvements (PSNR 24.15, SSIM 0.81) and trajectories that align with real data, signaling practical impact for infant neurodevelopment studies.

Abstract

Early infancy is a rapid and dynamic neurodevelopmental period for behavior and neurocognition. Longitudinal magnetic resonance imaging (MRI) is an effective tool to investigate such a crucial stage by capturing the developmental trajectories of the brain structures. However, longitudinal MRI acquisition always meets a serious data-missing problem due to participant dropout and failed scans, making longitudinal infant brain atlas construction and developmental trajectory delineation quite challenging. Thanks to the development of an AI-based generative model, neuroimage completion has become a powerful technique to retain as much available data as possible. However, current image completion methods usually suffer from inconsistency within each individual subject in the time dimension, compromising the overall quality. To solve this problem, our paper proposed a two-stage cascaded diffusion model, Cas-DiffCom, for dense and longitudinal 3D infant brain MRI completion and super-resolution. We applied our proposed method to the Baby Connectome Project (BCP) dataset. The experiment results validate that Cas-DiffCom achieves both individual consistency and high fidelity in longitudinal infant brain image completion. We further applied the generated infant brain images to two downstream tasks, brain tissue segmentation and developmental trajectory delineation, to declare its task-oriented potential in the neuroscience field.
Paper Structure (9 sections, 2 equations, 3 figures, 1 table)

This paper contains 9 sections, 2 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Generate Stage: Diffusion-based 3D Medical Image Completion
  4. Refine stage: Conditional 3D super-resolution
  5. Experiments and results
  6. Dataset and Image Processing
  7. Implementation Details
  8. Results
  9. Conclusion

Figures (3)

  • Figure 1: The denoising module in generate and fine stages. a) The cascaded diffusion pipeline for image completion consists of an image generation model and a super-resolution model. b) The asynchronous multimodal (AsMM) block. c) The super-resolution block with transposed convolution.
  • Figure 2: Showcase of the generated longitudinal infant brain images by our proposed Cas-DiffCom from the same subject. The top row displays the generated T1w MRI scans. The bottom row shows the corresponding brain segmentation results. The age time points are 3, 6, 9, 12, 20, and 26 months old. The image on the right at 24 months old is the real image from the same subject.
  • Figure 3: Comparison of brain tissue growth characterization between the generated data (red) and the ground truth (grey). The ground truth total volumes of the grey matter, white matter and cerebrospinal fluid fitted very well with the development trajectories estimated from the generated images.