CV-Attention UNet: Attention-based UNet for 3D Cerebrovascular Segmentation of Enhanced TOF-MRA Images

Syed Farhan Abbas,Nguyen Thanh Duc,Yoonguu Song,Kyungwon Kim,Ekta Srivastava,Boreom Lee

2024-06-19

Abstract:Due to the lack of automated methods, to diagnose cerebrovascular disease, time-of-flight magnetic resonance angiography (TOF-MRA) is assessed visually, making it time-consuming. The commonly used encoder-decoder architectures for cerebrovascular segmentation utilize redundant features, eventually leading to the extraction of low-level features multiple times. Additionally, convolutional neural networks (CNNs) suffer from performance degradation when the batch size is small, and deeper networks experience the vanishing gradient problem. Methods: In this paper, we attempt to solve these limitations and propose the 3D cerebrovascular attention UNet method, named CV-AttentionUNet, for precise extraction of brain vessel images. We proposed a sequence of preprocessing techniques followed by deeply supervised UNet to improve the accuracy of segmentation of the brain vessels leading to a stroke. To combine the low and high semantics, we applied the attention mechanism. This mechanism focuses on relevant associations and neglects irrelevant anatomical information. Furthermore, the inclusion of deep supervision incorporates different levels of features that prove to be beneficial for network convergence. Results: We demonstrate the efficiency of the proposed method by cross-validating with an unlabeled dataset, which was further labeled by us. We believe that the novelty of this algorithm lies in its ability to perform well on both labeled and unlabeled data with image processing-based enhancement. The results indicate that our method performed better than the existing state-of-the-art methods on the TubeTK dataset. Conclusion: The proposed method will help in accurate segmentation of cerebrovascular structure leading to stroke

Image and Video Processing,Computer Vision and Pattern Recognition,Machine Learning

What problem does this paper attempt to address?

The paper attempts to address several limitations of existing cerebrovascular segmentation methods when processing Time-of-Flight Magnetic Resonance Angiography (TOF-MRA) images. Specifically:

1. **Insufficient Automated Diagnosis**: Currently, TOF-MRA images are primarily used for visual assessment, lacking automated diagnostic methods, resulting in a time-consuming and inefficient diagnostic process.
2. **Redundant Feature Extraction**: Common encoder-decoder architectures in cerebrovascular segmentation repeatedly extract low-level features, leading to resource wastage and performance degradation.
3. **Decreased Performance in Small Batch Training**: Convolutional Neural Networks (CNNs) perform poorly in small batch training, and deep networks also encounter the vanishing gradient problem.
4. **Interference from Complex Background Structures**: The complex background structures and overlap between vessels in cerebrovascular images make the segmentation task more challenging.

To address these issues, the paper proposes a novel 3D cerebrovascular attention UNet method (CV-AttentionUNet), which improves the accuracy and efficiency of cerebrovascular segmentation by introducing attention mechanisms and deep supervision. The main contributions of this method include:

- **Preprocessing Techniques**: Utilizing a Hessian matrix-based vessel enhancement filter to improve vessel contrast and reduce interference from complex background structures.
- **Attention Mechanism**: Combining low-level and high-level semantic information, the attention mechanism focuses on relevant anatomical information while ignoring irrelevant parts.
- **Deep Supervision**: Integrating features from different levels through deep supervision helps the network converge and improves segmentation performance.

Experimental results show that this method outperforms existing state-of-the-art methods on the TubeTK dataset, especially in handling unannotated data. This research is expected to assist in accurately segmenting cerebrovascular structures in clinical applications, thereby aiding in the diagnosis and prognosis of stroke.