Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unsupervised Detection of Post-Stroke Brain Abnormalities

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

TL;DR

This paper presents REFLECT, a flow-based unsupervised framework for detecting post-stroke brain abnormalities beyond focal lesions. By training on healthy anatomy (IXI) rather than lesion-free stroke data (ATLAS), the model generalises better to diverse structural changes, including non-lesional abnormalities like atrophy and ventricular enlargement. Evaluation on ATLAS test slices shows improved anomaly sensitivity with IXI training, though lesion boundary precision may suffer, highlighting a trade-off between generalisation and contour accuracy. The work demonstrates the potential of reconstruction-based unsupervised methods to complement lesion-focused analyses and support longitudinal monitoring in stroke recovery, with avenues for 3D modelling and multimodal integration in the future.

Abstract

Post-stroke MRI not only delineates focal lesions but also reveals secondary structural changes, such as atrophy and ventricular enlargement. These abnormalities, increasingly recognised as imaging biomarkers of recovery and outcome, remain poorly captured by supervised segmentation methods. We evaluate REFLECT, a flow-based generative model, for unsupervised detection of both focal and non-lesional abnormalities in post-stroke patients. Using dual-expert central-slice annotations on ATLAS data, performance was assessed at the object level with Free-Response ROC analysis for anomaly maps. Two models were trained on lesion-free slices from stroke patients (ATLAS) and on healthy controls (IXI) to test the effect of training data. On ATLAS test subjects, the IXI-trained model achieved higher lesion segmentation (Dice = 0.37 vs 0.27) and improved sensitivity to non-lesional abnormalities (FROC = 0.62 vs 0.43). Training on fully healthy anatomy improves the modelling of normal variability, enabling broader and more reliable detection of structural abnormalities.

Unsupervised Detection of Post-Stroke Brain Abnormalities

TL;DR

This paper presents REFLECT, a flow-based unsupervised framework for detecting post-stroke brain abnormalities beyond focal lesions. By training on healthy anatomy (IXI) rather than lesion-free stroke data (ATLAS), the model generalises better to diverse structural changes, including non-lesional abnormalities like atrophy and ventricular enlargement. Evaluation on ATLAS test slices shows improved anomaly sensitivity with IXI training, though lesion boundary precision may suffer, highlighting a trade-off between generalisation and contour accuracy. The work demonstrates the potential of reconstruction-based unsupervised methods to complement lesion-focused analyses and support longitudinal monitoring in stroke recovery, with avenues for 3D modelling and multimodal integration in the future.

Abstract

Post-stroke MRI not only delineates focal lesions but also reveals secondary structural changes, such as atrophy and ventricular enlargement. These abnormalities, increasingly recognised as imaging biomarkers of recovery and outcome, remain poorly captured by supervised segmentation methods. We evaluate REFLECT, a flow-based generative model, for unsupervised detection of both focal and non-lesional abnormalities in post-stroke patients. Using dual-expert central-slice annotations on ATLAS data, performance was assessed at the object level with Free-Response ROC analysis for anomaly maps. Two models were trained on lesion-free slices from stroke patients (ATLAS) and on healthy controls (IXI) to test the effect of training data. On ATLAS test subjects, the IXI-trained model achieved higher lesion segmentation (Dice = 0.37 vs 0.27) and improved sensitivity to non-lesional abnormalities (FROC = 0.62 vs 0.43). Training on fully healthy anatomy improves the modelling of normal variability, enabling broader and more reliable detection of structural abnormalities.

Paper Structure

This paper contains 15 sections, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Data and Annotations
  4. Framework
  5. Training Protocol
  6. Evaluation
  7. Lesion detection and segmentation
  8. Anomaly detection in stroke patients
  9. Results
  10. Lesion segmentation and detection
  11. Detection of brain abnormalities
  12. Discussion
  13. Limitations
  14. Conclusion
  15. Acknowledgements

Figures (2)

  • Figure 1: FROC curves comparing ATLAS- and IXI-trained models across FPPI levels for overall, lesion and non-lesional abnormality detection at threshold $T=0.036$.
  • Figure 2: Examples of stroke-related anomaly detection from the ATLAS test set using the IXI-trained model. Top: pathological T1-w slices. Bottom, from left: segmented lesion (Orange outline = Ground Truth, Pink = Prediction), anomaly heatmaps with expert annotations ($\times$), illustrating localisation of non-lesional abnormalities, including enlarged sulci (middle) and periventricular hyposignals (right).