Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Meta-learning in healthcare: A survey

Alireza Rafiei, Ronald Moore, Sina Jahromi, Farshid Hajati, Rishikesan Kamaleswaran

TL;DR

This survey articulates how meta-learning—learning to learn—addresses core healthcare challenges such as data scarcity, heterogeneity, and distribution shifts. It categorizes methods into optimization-, metric-, and model-based families and maps them to healthcare tasks spanning clinical risk prediction, automated detection, drug discovery, and multi-/few-shot learning in imaging and text. The paper highlights advances (e.g., MAML and its variants, ProtoNets, MMANNs) and domain-specific applications (EHR, EEG, ECG, wearables, cancer imaging) while detailing practical considerations such as biases, validation, and computational costs. It concludes with a roadmap for broader adoption, emphasizing rigorous evaluation, interpretability, and integration with other learning paradigms to realize real-world impact in healthcare.

Abstract

As a subset of machine learning, meta-learning, or learning to learn, aims at improving the model's capabilities by employing prior knowledge and experience. A meta-learning paradigm can appropriately tackle the conventional challenges of traditional learning approaches, such as insufficient number of samples, domain shifts, and generalization. These unique characteristics position meta-learning as a suitable choice for developing influential solutions in various healthcare contexts, where the available data is often insufficient, and the data collection methodologies are different. This survey discusses meta-learning broad applications in the healthcare domain to provide insight into how and where it can address critical healthcare challenges. We first describe the theoretical foundations and pivotal methods of meta-learning. We then divide the employed meta-learning approaches in the healthcare domain into two main categories of multi/single-task learning and many/few-shot learning and survey the studies. Finally, we highlight the current challenges in meta-learning research, discuss the potential solutions, and provide future perspectives on meta-learning in healthcare.

Meta-learning in healthcare: A survey

TL;DR

This survey articulates how meta-learning—learning to learn—addresses core healthcare challenges such as data scarcity, heterogeneity, and distribution shifts. It categorizes methods into optimization-, metric-, and model-based families and maps them to healthcare tasks spanning clinical risk prediction, automated detection, drug discovery, and multi-/few-shot learning in imaging and text. The paper highlights advances (e.g., MAML and its variants, ProtoNets, MMANNs) and domain-specific applications (EHR, EEG, ECG, wearables, cancer imaging) while detailing practical considerations such as biases, validation, and computational costs. It concludes with a roadmap for broader adoption, emphasizing rigorous evaluation, interpretability, and integration with other learning paradigms to realize real-world impact in healthcare.

Abstract

As a subset of machine learning, meta-learning, or learning to learn, aims at improving the model's capabilities by employing prior knowledge and experience. A meta-learning paradigm can appropriately tackle the conventional challenges of traditional learning approaches, such as insufficient number of samples, domain shifts, and generalization. These unique characteristics position meta-learning as a suitable choice for developing influential solutions in various healthcare contexts, where the available data is often insufficient, and the data collection methodologies are different. This survey discusses meta-learning broad applications in the healthcare domain to provide insight into how and where it can address critical healthcare challenges. We first describe the theoretical foundations and pivotal methods of meta-learning. We then divide the employed meta-learning approaches in the healthcare domain into two main categories of multi/single-task learning and many/few-shot learning and survey the studies. Finally, we highlight the current challenges in meta-learning research, discuss the potential solutions, and provide future perspectives on meta-learning in healthcare.
Paper Structure (62 sections, 18 equations, 3 figures, 5 tables, 2 algorithms)

This paper contains 62 sections, 18 equations, 3 figures, 5 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Utilities of meta-learning in healthcare
  3. Fundamentals and key algorithms in meta-learning
  4. Optimization-based Models
  5. LSTM Meta-learner
  6. Model-agnostic Meta-learning
  7. First-order Model-Agnostic Meta-learning Extensions
  8. Second-order Model-Agnostic Meta-learning Extensions
  9. Metric-based Models
  10. Siamese neural networks
  11. Matching Networks
  12. Prototypical Networks
  13. Relation Networks
  14. Model-based Models
  15. Memory Augmented Neural Networks
  16. ...and 47 more sections

Figures (3)

  • Figure 1: The conceptual diagram of the fundamentals and key algorithms (left), and application domains of meta-learning in healthcare (right).
  • Figure 2: Rapid adaptation example of a meta-learning model to a new task by leveraging experience from various prior data sources in healthcare.
  • Figure 3: Example of many/few-shot medical image segmentation using meta-learning.