Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cross-Task Data Augmentation by Pseudo-label Generation for Region Based Coronary Artery Instance Segmentation

Sandesh Pokhrel, Sanjay Bhandari, Eduard Vazquez, Yash Raj Shrestha, Binod Bhattarai

TL;DR

The paper tackles the data scarcity problem in angiographic coronary artery segmentation by introducing cross-task data augmentation through pseudo-label generation on a related stenosis dataset. The authors train an intermediate model on the artery dataset, generate pseudo-labels for the stenosis data, and train a final YOLOv8-based segmentation model on a combined dataset, with enhancements such as CLAHE, unsharp masking, and a multi-term loss. The proposed Yolov8Pse approach outperforms several baselines including MaskDino and ConvNeXt variants, achieving a +9% improvement in validation F1 and a +3% improvement in test F1 on the ARCADE dataset, along with a high mAP/50 in validation. The results demonstrate that cross-task pseudo-labeling can effectively leverage related unlabeled data to improve medical image segmentation under data-scarce conditions, offering practical implications for reducing annotation burden in coronary artery analysis.

Abstract

Coronary Artery Diseases (CADs) although preventable, are one of the leading causes of death and disability. Diagnosis of these diseases is often difficult and resource intensive. Angiographic imaging segmentation of the arteries has evolved as a tool of assistance that helps clinicians make an accurate diagnosis. However, due to the limited amount of data and the difficulty in curating a dataset, the task of segmentation has proven challenging. In this study, we introduce the use of pseudo-labels to address the issue of limited data in the angiographic dataset to enhance the performance of the baseline YOLO model. Unlike existing data augmentation techniques that improve the model constrained to a fixed dataset, we introduce the use of pseudo-labels generated on a dataset of separate related task to diversify and improve model performance. This method increases the baseline F1 score by 9% in the validation data set and by 3% in the test data set.

Cross-Task Data Augmentation by Pseudo-label Generation for Region Based Coronary Artery Instance Segmentation

TL;DR

The paper tackles the data scarcity problem in angiographic coronary artery segmentation by introducing cross-task data augmentation through pseudo-label generation on a related stenosis dataset. The authors train an intermediate model on the artery dataset, generate pseudo-labels for the stenosis data, and train a final YOLOv8-based segmentation model on a combined dataset, with enhancements such as CLAHE, unsharp masking, and a multi-term loss. The proposed Yolov8Pse approach outperforms several baselines including MaskDino and ConvNeXt variants, achieving a +9% improvement in validation F1 and a +3% improvement in test F1 on the ARCADE dataset, along with a high mAP/50 in validation. The results demonstrate that cross-task pseudo-labeling can effectively leverage related unlabeled data to improve medical image segmentation under data-scarce conditions, offering practical implications for reducing annotation burden in coronary artery analysis.

Abstract

Coronary Artery Diseases (CADs) although preventable, are one of the leading causes of death and disability. Diagnosis of these diseases is often difficult and resource intensive. Angiographic imaging segmentation of the arteries has evolved as a tool of assistance that helps clinicians make an accurate diagnosis. However, due to the limited amount of data and the difficulty in curating a dataset, the task of segmentation has proven challenging. In this study, we introduce the use of pseudo-labels to address the issue of limited data in the angiographic dataset to enhance the performance of the baseline YOLO model. Unlike existing data augmentation techniques that improve the model constrained to a fixed dataset, we introduce the use of pseudo-labels generated on a dataset of separate related task to diversify and improve model performance. This method increases the baseline F1 score by 9% in the validation data set and by 3% in the test data set.
Paper Structure (14 sections, 2 equations, 2 figures, 3 tables)

This paper contains 14 sections, 2 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Methods
  4. Problem Formulation
  5. Cross-Task Dataset Pseudo-labelling
  6. Loss Function
  7. Experiments
  8. Dataset and Preprocessing
  9. Implementation details
  10. Quantitative Results
  11. Qualitative Results
  12. Ablation Studies
  13. Conclusion
  14. Acknowledgments:

Figures (2)

  • Figure 1: a) Training of the intermediate model on the original artery segmentation dataset b) Generating pseudo-labels for the unlabeled stenosis detection dataset c) Training the model on a combination of the original artery segmentation dataset and the resultant pseudo-labeled dataset.
  • Figure 2: Qualitative instance segmentation results on Vessel segmenation. Ground truth masks followed by the instance segmentation masks generated by ConvnextV2, Yolov8, MaskDino and Pseudolabel-Yolo(Yolov8Pse) are shown in the figure respectively.