Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Confident Pseudo-labeled Diffusion Augmentation for Canine Cardiomegaly Detection

Shiman Zhang, Lakshmikar Reddy Polamreddy, Youshan Zhang

TL;DR

The paper addresses the challenge of detecting canine cardiomegaly from X-ray images with limited labeled data. It introduces Confident Pseudo-labeled Diffusion Augmentation (CDA), a framework that generates synthetic X-ray images via diffusion models and assigns VHS-related labels, while using Monte Carlo Dropout-based pseudo-labeling to select high-confidence unlabeled samples for training. A composite loss and VHS computation scheme promotes accurate localization of key points and VHS-based categorization into three clinically meaningful classes. The results show state-of-the-art VHS prediction accuracy on a canine chest X-ray dataset, with 92.8% test accuracy, demonstrating improved generalization through synthetic augmentation and selective semi-supervised learning. The work has practical implications for scalable, robust canine cardiomegaly diagnosis in veterinary practice.

Abstract

Canine cardiomegaly, marked by an enlarged heart, poses serious health risks if undetected, requiring accurate diagnostic methods. Current detection models often rely on small, poorly annotated datasets and struggle to generalize across diverse imaging conditions, limiting their real-world applicability. To address these issues, we propose a Confident Pseudo-labeled Diffusion Augmentation (CDA) model for identifying canine cardiomegaly. Our approach addresses the challenge of limited high-quality training data by employing diffusion models to generate synthetic X-ray images and annotate Vertebral Heart Score key points, thereby expanding the dataset. We also employ a pseudo-labeling strategy with Monte Carlo Dropout to select high-confidence labels, refine the synthetic dataset, and improve accuracy. Iteratively incorporating these labels enhances the model's performance, overcoming the limitations of existing approaches. Experimental results show that the CDA model outperforms traditional methods, achieving state-of-the-art accuracy in canine cardiomegaly detection. The code implementation is available at https://github.com/Shira7z/CDA.

Confident Pseudo-labeled Diffusion Augmentation for Canine Cardiomegaly Detection

TL;DR

The paper addresses the challenge of detecting canine cardiomegaly from X-ray images with limited labeled data. It introduces Confident Pseudo-labeled Diffusion Augmentation (CDA), a framework that generates synthetic X-ray images via diffusion models and assigns VHS-related labels, while using Monte Carlo Dropout-based pseudo-labeling to select high-confidence unlabeled samples for training. A composite loss and VHS computation scheme promotes accurate localization of key points and VHS-based categorization into three clinically meaningful classes. The results show state-of-the-art VHS prediction accuracy on a canine chest X-ray dataset, with 92.8% test accuracy, demonstrating improved generalization through synthetic augmentation and selective semi-supervised learning. The work has practical implications for scalable, robust canine cardiomegaly diagnosis in veterinary practice.

Abstract

Canine cardiomegaly, marked by an enlarged heart, poses serious health risks if undetected, requiring accurate diagnostic methods. Current detection models often rely on small, poorly annotated datasets and struggle to generalize across diverse imaging conditions, limiting their real-world applicability. To address these issues, we propose a Confident Pseudo-labeled Diffusion Augmentation (CDA) model for identifying canine cardiomegaly. Our approach addresses the challenge of limited high-quality training data by employing diffusion models to generate synthetic X-ray images and annotate Vertebral Heart Score key points, thereby expanding the dataset. We also employ a pseudo-labeling strategy with Monte Carlo Dropout to select high-confidence labels, refine the synthetic dataset, and improve accuracy. Iteratively incorporating these labels enhances the model's performance, overcoming the limitations of existing approaches. Experimental results show that the CDA model outperforms traditional methods, achieving state-of-the-art accuracy in canine cardiomegaly detection. The code implementation is available at https://github.com/Shira7z/CDA.
Paper Structure (16 sections, 6 equations, 4 figures, 1 table, 2 algorithms)

This paper contains 16 sections, 6 equations, 4 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Confident Pseudo-labeled Diffusion Augmentation (CDA) model
  5. Loss Function.
  6. VHS Calculation.
  7. Pseudo-Labeling with Monte Carlo Dropout
  8. Loss Function for Pseudo-Labeling
  9. Datasets
  10. Data Collection
  11. Data Labeling
  12. Implementation details
  13. Results
  14. Ablation study
  15. Discussion
  16. ...and 1 more sections

Figures (4)

  • Figure 1: Overall framework of the Confident Pseudo-labeled Diffusion Augmentation (CDA) model. The framework integrates synthetic data augmentation, high-confidence pseudo-labeling, and iterative training for reliable VHS prediction.
  • Figure 2: Illustration of anatomical landmarks for VHS calculation. $AB$ (long axis), $CD$ (short axis), and $EF$ (vertebral segment) are shown, which are used to compute the Vertebral Heart Score (VHS).
  • Figure 3: Comparisions of high confidence prediction and low confidence presiction.
  • Figure 4: Prediction comparisons of different models.