Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unsupervised Deep Generative Models for Anomaly Detection in Neuroimaging: A Systematic Scoping Review

Youwan Mahé, Elise Bannier, Stéphanie Leplaideur, Elisa Fromont, Francesca Galassi

TL;DR

This systematic scoping review surveys unsupervised deep generative models for brain anomaly detection across 2018–2025, covering autoencoders, VAEs, GANs, diffusion models, and related non-generative approaches. It highlights that large, focal tumours remain the most tractable target, while small or sparse lesions such as MS, WMH, and stroke pose consistent challenges; 3D architectures and advanced losses improve performance but rarely reach supervised baselines. A key strength of these methods is the ability to produce pseudo-healthy counterfactual reconstructions, offering interpretable visual explanations in data-scarce settings and enabling within- and cross-disease deviation mapping. The review identifies core directions for clinical translation, including anatomy-aware modelling, foundation models, standardized cross-pathology benchmarks, and prospective reader studies to validate whether counterfactual reconstructions improve diagnostic decision-making.

Abstract

Unsupervised deep generative models are emerging as a promising alternative to supervised methods for detecting and segmenting anomalies in brain imaging. Unlike fully supervised approaches, which require large voxel-level annotated datasets and are limited to well-characterised pathologies, these models can be trained exclusively on healthy data and identify anomalies as deviations from learned normative brain structures. This PRISMA-guided scoping review synthesises recent work on unsupervised deep generative models for anomaly detection in neuroimaging, including autoencoders, variational autoencoders, generative adversarial networks, and denoising diffusion models. A total of 49 studies published between 2018 - 2025 were identified, covering applications to brain MRI and, less frequently, CT across diverse pathologies such as tumours, stroke, multiple sclerosis, and small vessel disease. Reported performance metrics are compared alongside architectural design choices. Across the included studies, generative models achieved encouraging performance for large focal lesions and demonstrated progress in addressing more subtle abnormalities. A key strength of generative models is their ability to produce interpretable pseudo-healthy (also referred to as counterfactual) reconstructions, which is particularly valuable when annotated data are scarce, as in rare or heterogeneous diseases. Looking ahead, these models offer a compelling direction for anomaly detection, enabling semi-supervised learning, supporting the discovery of novel imaging biomarkers, and facilitating within- and cross-disease deviation mapping in unified end-to-end frameworks. To realise clinical impact, future work should prioritise anatomy-aware modelling, development of foundation models, task-appropriate evaluation metrics, and rigorous clinical validation.

Unsupervised Deep Generative Models for Anomaly Detection in Neuroimaging: A Systematic Scoping Review

TL;DR

This systematic scoping review surveys unsupervised deep generative models for brain anomaly detection across 2018–2025, covering autoencoders, VAEs, GANs, diffusion models, and related non-generative approaches. It highlights that large, focal tumours remain the most tractable target, while small or sparse lesions such as MS, WMH, and stroke pose consistent challenges; 3D architectures and advanced losses improve performance but rarely reach supervised baselines. A key strength of these methods is the ability to produce pseudo-healthy counterfactual reconstructions, offering interpretable visual explanations in data-scarce settings and enabling within- and cross-disease deviation mapping. The review identifies core directions for clinical translation, including anatomy-aware modelling, foundation models, standardized cross-pathology benchmarks, and prospective reader studies to validate whether counterfactual reconstructions improve diagnostic decision-making.

Abstract

Unsupervised deep generative models are emerging as a promising alternative to supervised methods for detecting and segmenting anomalies in brain imaging. Unlike fully supervised approaches, which require large voxel-level annotated datasets and are limited to well-characterised pathologies, these models can be trained exclusively on healthy data and identify anomalies as deviations from learned normative brain structures. This PRISMA-guided scoping review synthesises recent work on unsupervised deep generative models for anomaly detection in neuroimaging, including autoencoders, variational autoencoders, generative adversarial networks, and denoising diffusion models. A total of 49 studies published between 2018 - 2025 were identified, covering applications to brain MRI and, less frequently, CT across diverse pathologies such as tumours, stroke, multiple sclerosis, and small vessel disease. Reported performance metrics are compared alongside architectural design choices. Across the included studies, generative models achieved encouraging performance for large focal lesions and demonstrated progress in addressing more subtle abnormalities. A key strength of generative models is their ability to produce interpretable pseudo-healthy (also referred to as counterfactual) reconstructions, which is particularly valuable when annotated data are scarce, as in rare or heterogeneous diseases. Looking ahead, these models offer a compelling direction for anomaly detection, enabling semi-supervised learning, supporting the discovery of novel imaging biomarkers, and facilitating within- and cross-disease deviation mapping in unified end-to-end frameworks. To realise clinical impact, future work should prioritise anatomy-aware modelling, development of foundation models, task-appropriate evaluation metrics, and rigorous clinical validation.
Paper Structure (57 sections, 4 equations, 3 figures, 1 table)

This paper contains 57 sections, 4 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Review questions
  3. Methods
  4. Background: generative modelling for UDA
  5. Information sources and search strategy
  6. Eligibility and screening
  7. Risk of bias assessment
  8. Results
  9. Study selection
  10. Study characteristics
  11. Evaluation metrics
  12. Datasets and pathologies
  13. Brain tumours.
  14. Stroke.
  15. Multiple sclerosis and white-matter hyperintensities.
  16. ...and 42 more sections

Figures (3)

  • Figure 1: PRISMA flow diagram for scoping reviews, including database and register searches.
  • Figure 2: Central axial slices from a healthy brain (IXI), a brain tumour case (BraTS), a chronic stroke case (ATLAS v2.0), and a multiple sclerosis case (MSSEG).
  • Figure 3: Mean Dice scores ($\pm$ SD) of unsupervised anomaly detection methods across pathologies. Reported values are derived from different datasets, preprocessing pipelines, and evaluation protocols; therefore, absolute values are not directly comparable between families and should be interpreted as indicative trends rather than head-to-head benchmarks.