Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Weakly supervised segmentation of intracranial aneurysms using a novel 3D focal modulation UNet

Amirhossein Rasoulian, Arash Harirpoush, Soorena Salari, Yiming Xiao

TL;DR

The paper addresses automatic detection and segmentation of unruptured intracranial aneurysms (UIAs) from TOF-MRA under weak supervision. It introduces FocalSegNet, a 3D UNet-like architecture that employs focal modulation in the encoder to capture contextual information more efficiently than self-attention, and pairs it with a fully connected CRF post-processing stage to refine coarse predictions into accurate voxel-wise segmentations. On a public TOF-MRA UIA dataset, FocalSegNet achieves a Dice score of $0.677 \pm 0.141$ and a 95-HD of $2.15$ voxels, with a false-positive rate of $0.212$ and sensitivity of $0.801$, outperforming the 3D Residual-UNet and approaching Swin-UNETR performance. The combination of weak labeling, focal modulation, and CRF refinement provides a practical and efficient pipeline for UIA analysis with potential impact on risk assessment and treatment planning in cerebrovascular disease.

Abstract

Accurate identification and quantification of unruptured intracranial aneurysms (UIAs) is crucial for the risk assessment and treatment of this cerebrovascular disorder. Current 2D manual assessment on 3D magnetic resonance angiography (MRA) is suboptimal and time-consuming. In addition, one major issue in medical image segmentation is the need for large well-annotated data, which can be expensive to obtain. Techniques that mitigate this requirement, such as weakly supervised learning with coarse labels are highly desirable. In the paper, we propose FocalSegNet, a novel 3D focal modulation UNet, to detect an aneurysm and offer an initial, coarse segmentation of it from time-of-flight MRA image patches, which is further refined with a dense conditional random field (CRF) post-processing layer to produce a final segmentation map. We trained and evaluated our model on a public dataset, and in terms of UIA detection, our model showed a low false-positive rate of 0.21 and a high sensitivity of 0.80. For voxel-wise aneurysm segmentation, we achieved a Dice score of 0.68 and a 95% Hausdorff distance of ~0.95 mm, demonstrating its strong performance. We evaluated our algorithms against the state-of-the-art 3D Residual-UNet and Swin-UNETR, and illustrated the superior performance of our proposed FocalSegNet, highlighting the advantages of employing focal modulation for this task.

Weakly supervised segmentation of intracranial aneurysms using a novel 3D focal modulation UNet

TL;DR

The paper addresses automatic detection and segmentation of unruptured intracranial aneurysms (UIAs) from TOF-MRA under weak supervision. It introduces FocalSegNet, a 3D UNet-like architecture that employs focal modulation in the encoder to capture contextual information more efficiently than self-attention, and pairs it with a fully connected CRF post-processing stage to refine coarse predictions into accurate voxel-wise segmentations. On a public TOF-MRA UIA dataset, FocalSegNet achieves a Dice score of and a 95-HD of voxels, with a false-positive rate of and sensitivity of , outperforming the 3D Residual-UNet and approaching Swin-UNETR performance. The combination of weak labeling, focal modulation, and CRF refinement provides a practical and efficient pipeline for UIA analysis with potential impact on risk assessment and treatment planning in cerebrovascular disease.

Abstract

Accurate identification and quantification of unruptured intracranial aneurysms (UIAs) is crucial for the risk assessment and treatment of this cerebrovascular disorder. Current 2D manual assessment on 3D magnetic resonance angiography (MRA) is suboptimal and time-consuming. In addition, one major issue in medical image segmentation is the need for large well-annotated data, which can be expensive to obtain. Techniques that mitigate this requirement, such as weakly supervised learning with coarse labels are highly desirable. In the paper, we propose FocalSegNet, a novel 3D focal modulation UNet, to detect an aneurysm and offer an initial, coarse segmentation of it from time-of-flight MRA image patches, which is further refined with a dense conditional random field (CRF) post-processing layer to produce a final segmentation map. We trained and evaluated our model on a public dataset, and in terms of UIA detection, our model showed a low false-positive rate of 0.21 and a high sensitivity of 0.80. For voxel-wise aneurysm segmentation, we achieved a Dice score of 0.68 and a 95% Hausdorff distance of ~0.95 mm, demonstrating its strong performance. We evaluated our algorithms against the state-of-the-art 3D Residual-UNet and Swin-UNETR, and illustrated the superior performance of our proposed FocalSegNet, highlighting the advantages of employing focal modulation for this task.
Paper Structure (15 sections, 8 equations, 2 figures, 2 tables)

This paper contains 15 sections, 8 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Network architecture of FocalSegNet
  4. Post-processing with fully connected CRF
  5. Experiments and Evaluation
  6. Dataset and pre-processing
  7. Implementation details and baseline models
  8. Sampling and data augmentation:
  9. Loss function
  10. Evaluation metrics
  11. Results
  12. UIA detection and segmentation
  13. Ablation studies
  14. Discussion
  15. Conclusion

Figures (2)

  • Figure 1: Network architecture of the proposed FocalSegNet
  • Figure 2: Comparison of segmentation results of different techniques for two different patients (one patient per row).Red label=ground truths and green label=automatic segmentation.