Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DuMeta++: Spatiotemporal Dual Meta-Learning for Generalizable Few-Shot Brain Tissue Segmentation Across Diverse Ages

Yongheng Sun, Jun Shu, Jianhua Ma, Fan Wang

TL;DR

DuMeta++ tackles cross-age generalization in brain tissue segmentation under few-shot scenarios without relying on paired longitudinal scans. It combines meta-feature learning to yield an age-invariant encoder with meta-initialization learning for rapid segmentation-head adaptation, reinforced by a memory-bank based class-aware regularization that enforces longitudinal consistency across ages. The framework comes with theoretical convergence guarantees for its dual bilevel optimization and demonstrates superior cross-age performance on iSeg-2019, IBIS, OASIS, and ADNI in one- and five-shot settings, along with improved longitudinal stability. This approach offers a practical, data-efficient solution for lifespan neuroimaging that reduces reliance on longitudinal data and extensive labeled annotations.

Abstract

Accurate segmentation of brain tissues from MRI scans is critical for neuroscience and clinical applications, but achieving consistent performance across the human lifespan remains challenging due to dynamic, age-related changes in brain appearance and morphology. While prior work has sought to mitigate these shifts by using self-supervised regularization with paired longitudinal data, such data are often unavailable in practice. To address this, we propose \emph{DuMeta++}, a dual meta-learning framework that operates without paired longitudinal data. Our approach integrates: (1) meta-feature learning to extract age-agnostic semantic representations of spatiotemporally evolving brain structures, and (2) meta-initialization learning to enable data-efficient adaptation of the segmentation model. Furthermore, we propose a memory-bank-based class-aware regularization strategy to enforce longitudinal consistency without explicit longitudinal supervision. We theoretically prove the convergence of our DuMeta++, ensuring stability. Experiments on diverse datasets (iSeg-2019, IBIS, OASIS, ADNI) under few-shot settings demonstrate that DuMeta++ outperforms existing methods in cross-age generalization. Code will be available at https://github.com/ladderlab-xjtu/DuMeta++.

DuMeta++: Spatiotemporal Dual Meta-Learning for Generalizable Few-Shot Brain Tissue Segmentation Across Diverse Ages

TL;DR

DuMeta++ tackles cross-age generalization in brain tissue segmentation under few-shot scenarios without relying on paired longitudinal scans. It combines meta-feature learning to yield an age-invariant encoder with meta-initialization learning for rapid segmentation-head adaptation, reinforced by a memory-bank based class-aware regularization that enforces longitudinal consistency across ages. The framework comes with theoretical convergence guarantees for its dual bilevel optimization and demonstrates superior cross-age performance on iSeg-2019, IBIS, OASIS, and ADNI in one- and five-shot settings, along with improved longitudinal stability. This approach offers a practical, data-efficient solution for lifespan neuroimaging that reduces reliance on longitudinal data and extensive labeled annotations.

Abstract

Accurate segmentation of brain tissues from MRI scans is critical for neuroscience and clinical applications, but achieving consistent performance across the human lifespan remains challenging due to dynamic, age-related changes in brain appearance and morphology. While prior work has sought to mitigate these shifts by using self-supervised regularization with paired longitudinal data, such data are often unavailable in practice. To address this, we propose \emph{DuMeta++}, a dual meta-learning framework that operates without paired longitudinal data. Our approach integrates: (1) meta-feature learning to extract age-agnostic semantic representations of spatiotemporally evolving brain structures, and (2) meta-initialization learning to enable data-efficient adaptation of the segmentation model. Furthermore, we propose a memory-bank-based class-aware regularization strategy to enforce longitudinal consistency without explicit longitudinal supervision. We theoretically prove the convergence of our DuMeta++, ensuring stability. Experiments on diverse datasets (iSeg-2019, IBIS, OASIS, ADNI) under few-shot settings demonstrate that DuMeta++ outperforms existing methods in cross-age generalization. Code will be available at https://github.com/ladderlab-xjtu/DuMeta++.
Paper Structure (31 sections, 4 theorems, 47 equations, 8 figures, 7 tables, 1 algorithm)

This paper contains 31 sections, 4 theorems, 47 equations, 8 figures, 7 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Brain Tissue Segmentation
  4. Meta-Feature Learning
  5. Meta-Initialization Learning
  6. Longitudinally-Consistent Representation Learning
  7. Method
  8. DuMeta++ Paradigm
  9. Meta-Learning of the Universal Feature Extractor
  10. Meta-Learning of the Initialized Segmentation Head
  11. Class-Aware Regularization
  12. Convergence of the Dual Meta-Learning Pipeline
  13. Experiment
  14. Experimental Setup
  15. Datasets
  16. ...and 16 more sections

Key Result

Theorem 1

Suppose the loss function $L_{\text{outer1}}$ is Lipschitz smooth with constant $L$, and the gradient of $\theta$ with respect to the loss function $L_{\text{outer1}}$ is Lipschitz continuous with constant $L$. Let the learning rate $\alpha_t, \beta_t, 1\leq t\leq T$ be monotonically descent sequenc where $C$ is a constant independent of the convergence process.

Figures (8)

  • Figure 1: (a) Brain morphology and tissue contrast of 6-month-old infants from the iSeg-2019 dataset. (b) Aging and Alzheimer's processes of the elderly from the ADNI dataset.
  • Figure 2: Schematic diagram of DuMeta++. The meta-training process is divided into three steps, leveraging a shared inner loop but leading to two distinct outer loops for learning a feature extractor and segmentation head. Specifically, steps (1) and (2) comprise the MFL procedure for the frozen feature extractor, while steps (1) and (3) realize the MIL procedure for establishing a well-initialized segmentation head. At inference time, we fine-tune only the segmentation head using data from the new, unseen domain.
  • Figure 3: The 2D slice views of representative one-shot segmentation results on the held-out test set of iSeg-2019.
  • Figure 4: The 2D slice views of representative one-shot segmentation results on the held-out test set of ADNI.
  • Figure 5: Visualization of segmentation results of the same subject at multiple time points on ADNI.
  • ...and 3 more figures

Theorems & Definitions (6)

  • Theorem 1
  • Theorem 2
  • Theorem 3
  • proof
  • Theorem 4
  • proof