OCTAve: 2D en face Optical Coherence Tomography Angiography Vessel Segmentation in Weakly-Supervised Learning with Locality Augmentation

Amrest Chinkamol, Vetit Kanjaras, Phattarapong Sawangjai, Yitian Zhao, Thapanun Sudhawiyangkul, Chantana Chantrapornchai, Cuntai Guan, Theerawit Wilaiprasitporn

Read on ArXiv https://arxiv.org/abs/2207.12238

Read on IEEE Transactions on Biomedical Engineering https://doi.org/10.1109/TBME.2022.3232102

architecture-figure

TL;DR

In this work, we tried to lessen the expert clinician works on the segmentation labelling task of 2D en face OCTA images by adopting the scribble-like ground truth training technique proposed by Valvano et al. Which is working to some degree, but we found some problems with the original technique such as performance issue and segmentation artifacts. To solve these issues, we proposed a new deep-supervision technique called Self-Supervised Deep Supervision (SSDS) to enhance the model learning and suppress segmentation artifacts in the prediction results. In our experiment, we found that SSDS significantly improve not just weakly-supervised learning from scribble-like ground truth, but also a generic fully-supervised learning as well.

Abstract

Objective:

While the microvasculature annotation within Optical Coherence Tomography Angiography (OCTA) can be leveraged using deep-learning techniques, expensive annotation processes are required to create sufficient training data. One way to avoid the expensive annotation is to use a type of weak annotation in which only the center of the vessel is annotated. However, retaining the final segmentation quality with roughly annotated data remains a challenge.

Methods:

Our proposed methods called OCTAve provide a new way of using weak-annotation on microvasculature segmentation. Since the centerline labels are similar to scribble annotations, we attempted to solve this problem by using the scribble-based weakly-supervised learning method. Even though the initial results look promising, we found that the method could be significantly improved by adding our novel self-supervised deep supervision method based on Kullback-Liebler divergence.

Results:

The study on large public datasets with different annotation styles (i.e., ROSE, OCTA-500) demonstrates that our proposed method gives better quantitative and qualitative results than the baseline methods and a naive approach, with a p-value less than 0.001 on dice-coefficients and a lot fewer artifacts visually seen.

Conclusion:

The segmentation results are both qualitatively and quantitatively superior to baseline weakly-supervised methods when using scribble-based weakly-supervised learning augmented with self-supervised deep supervision, with an average drop in segmentation performance of less than 10%. Significance: This work gives a new perspective on how weakly-supervised learning can be used to reduce the cost of annotating microvasculature, which can make the annotating process easier and reduce the amount of work for domain experts.

Keywords

Optical Coherence Tomography Angiography, Vessel Segmentation, Deep Neural Network, Self-Supervised Learning, Weakly-Supervised Learning

Citation

@ARTICLE{9999313,
    author={Chinkamol, Amrest and Kanjaras, Vetit and Sawangjai, Phattarapong and Zhao, Yitian and Sudhawiyangkul, Thapanun and Chantrapornchai, Chantana and Guan, Cuntai and Wilaiprasitporn, Theerawit},
    journal={IEEE Transactions on Biomedical Engineering},
    title={OCTAve: 2D en face Optical Coherence Tomography Angiography Vessel Segmentation in Weakly-Supervised Learning with Locality Augmentation},
    year={2022},
    volume={},
    number={},
    pages={1-12},
    doi={10.1109/TBME.2022.3232102}
}