Domain Adaptation and Generalization on Functional Medical Images: A Systematic Survey

Gita Sarafraz, Armin Behnamnia, Mehran Hosseinzadeh, Ali Balapour, Amin Meghrazi, Hamid R. RabieearXiv 2022 · DOI · arXiv

Introduction: The Challenge of Distribution Shift

Deep learning models typically rely on the independent and identically distributed (i.i.d.) assumption—the belief that training and testing data share the same underlying distribution. In functional brain imaging (e.g., EEG, fMRI), this assumption is frequently violated due to "domain shift." Variations in hardware, recording environments, and individual subject physiology lead to significant performance degradation when models are deployed in real-world clinical settings. This systematic survey explores strategies to bridge this gap through Domain Adaptation (DA) and Domain Generalization (DG).

A Taxonomy of Transfer Learning and Adaptation

The field is broadly categorized based on the relationship between source and target domains, specifically regarding marginal distributions $P(X)$ and label spaces $Y$. Understanding these relationships is crucial for selecting the appropriate transfer strategy.

Taxonomy of Transfer Learning and Domain Adaptation settings based on distribution and label shifts.

Domain Adaptation (DA)

DA methods assume access to some target domain data (often unlabeled) during training. Key strategies include:

  • Alignment: Techniques like Domain-Adversarial Neural Networks (DANN) use adversarial training to force the model to learn domain-invariant features.
  • Feature Disentanglement: Separating task-relevant brain signals from domain-specific noise.
  • Pseudo-labeling: Iteratively refining model predictions on the target domain to bootstrap learning.

Domain Generalization (DG)

DG aims for "zero-shot" transfer to unseen domains by learning universal representations that work out-of-the-box. Strategies include:

  • Meta-Learning: Simulating domain shifts during training to "learn to learn" across environments.
  • Data Augmentation: Synthesizing diverse domains to improve robustness to noise.
  • Ensemble Learning: Leveraging multiple source models to mitigate individual site biases.

Scenarios in Domain Adaptation

Beyond simple distribution shifts, the overlap between source and target label sets defines different DA challenges. While Closed Set DA assumes identical classes, more complex scenarios like Universal DA handle unknown label relationships.

Comparison of Universal DA and Different Set DA scenarios, illustrating label set complexity.

Architectures and Benchmarks

The survey highlights a transition from traditional machine learning (SVMs) to deep architectures:

  • CNNs: The most widely used building block, appearing in over 50% of fMRI studies reviewed.
  • RNNs/LSTMs: Frequently employed to capture temporal sequences in EEG data.
  • Graph Neural Networks (GNNs): Rapidly gaining ground for their ability to model the physical layout of sensors and functional brain connectivity.

Standard benchmarks include SEED and DEAP for EEG emotion recognition, and ABIDE (Autism Brain Imaging Data Exchange) for fMRI-based diagnostic generalization.

Key Insights and Future Directions

Functional images are uniquely sensitive to non-stationary noise (e.g., subject mood, caffeine intake, or electrode displacement), making "Cross-Subject" adaptation the most significant challenge. While adversarial alignment is currently the most popular strategy, there is a clear trend toward Unsupervised Domain Adaptation to circumvent the high cost of medical data labeling.

Future research must address several "frontier" areas:

  • Interpretability: Moving beyond "black box" models to provide clinical justifications for predictions.
  • Privacy-Preserving DA: Utilizing Federated Learning to adapt models without exposing sensitive patient data.
  • Source-Free Adaptation: Developing methods that can adapt a pre-trained model to a new site without requiring access to the original training data.

Summary updated

Abstract

Machine learning algorithms have revolutionized different fields, including natural language processing, computer vision, signal processing, and medical data processing. Despite the excellent capabilities of machine learning algorithms in various tasks and areas, the performance of these models mainly deteriorates when there is a shift in the test and training data distributions. This gap occurs due to the violation of the fundamental assumption that the training and test data are independent and identically distributed (i.i.d). In real-world scenarios where collecting data from all possible domains for training is costly and even impossible, the i.i.d assumption can hardly be satisfied. The problem is even more severe in the case of medical images and signals because it requires either expensive equipment or a meticulous experimentation setup to collect data, even for a single domain. Additionally, the decrease in performance may have severe consequences in the analysis of medical records. As a result of such problems, the ability to generalize and adapt under distribution shifts (domain generalization (DG) and domain adaptation (DA)) is essential for the analysis of medical data. This paper provides the first systematic review of DG and DA on functional brain signals to fill the gap of the absence of a comprehensive study in this era. We provide detailed explanations and categorizations of datasets, approaches, and architectures used in DG and DA on functional brain images. We further address the attention-worthy future tracks in this field.
BibTeX
@misc{sarafraz2022domain,
  author = {Gita Sarafraz and Armin Behnamnia and Mehran Hosseinzadeh and Ali Balapour and Amin Meghrazi and Hamid R. Rabiee},
  title = {{Domain Adaptation and Generalization on Functional Medical Images: A Systematic Survey}},
  year = {2022},
  eprint = {2212.03176},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2212.03176},
}
Figure