Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Transesophageal Echocardiography Generation using Anatomical Models

Emmanuel Oladokun, Musa Abdulkareem, Jurica Šprem, Vicente Grau

TL;DR

A pipeline to generate synthetic TEE images and corresponding semantic labels is developed and it is demonstrated that such images can improve DL network performance through a left-ventricle semantic segmentation task.

Abstract

Through automation, deep learning (DL) can enhance the analysis of transesophageal echocardiography (TEE) images. However, DL methods require large amounts of high-quality data to produce accurate results, which is difficult to satisfy. Data augmentation is commonly used to tackle this issue. In this work, we develop a pipeline to generate synthetic TEE images and corresponding semantic labels. The proposed data generation pipeline expands on an existing pipeline that generates synthetic transthoracic echocardiography images by transforming slices from anatomical models into synthetic images. We also demonstrate that such images can improve DL network performance through a left-ventricle semantic segmentation task. For the pipeline's unpaired image-to-image (I2I) translation section, we explore two generative methods: CycleGAN and contrastive unpaired translation. Next, we evaluate the synthetic images quantitatively using the Fréchet Inception Distance (FID) Score and qualitatively through a human perception quiz involving expert cardiologists and the average researcher. In this study, we achieve a dice score improvement of up to 10% when we augment datasets with our synthetic images. Furthermore, we compare established methods of assessing unpaired I2I translation and observe a disagreement when evaluating the synthetic images. Finally, we see which metric better predicts the generated data's efficacy when used for data augmentation.

Transesophageal Echocardiography Generation using Anatomical Models

TL;DR

A pipeline to generate synthetic TEE images and corresponding semantic labels is developed and it is demonstrated that such images can improve DL network performance through a left-ventricle semantic segmentation task.

Abstract

Through automation, deep learning (DL) can enhance the analysis of transesophageal echocardiography (TEE) images. However, DL methods require large amounts of high-quality data to produce accurate results, which is difficult to satisfy. Data augmentation is commonly used to tackle this issue. In this work, we develop a pipeline to generate synthetic TEE images and corresponding semantic labels. The proposed data generation pipeline expands on an existing pipeline that generates synthetic transthoracic echocardiography images by transforming slices from anatomical models into synthetic images. We also demonstrate that such images can improve DL network performance through a left-ventricle semantic segmentation task. For the pipeline's unpaired image-to-image (I2I) translation section, we explore two generative methods: CycleGAN and contrastive unpaired translation. Next, we evaluate the synthetic images quantitatively using the Fréchet Inception Distance (FID) Score and qualitatively through a human perception quiz involving expert cardiologists and the average researcher. In this study, we achieve a dice score improvement of up to 10% when we augment datasets with our synthetic images. Furthermore, we compare established methods of assessing unpaired I2I translation and observe a disagreement when evaluating the synthetic images. Finally, we see which metric better predicts the generated data's efficacy when used for data augmentation.

Paper Structure

This paper contains 15 sections, 2 equations, 1 figure, 3 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Pseudo-image Generation
  4. Anatomical Models
  5. TEE View Extraction
  6. Image Synthesis
  7. Unpaired I2I Models
  8. Data
  9. Generative Training
  10. Evaluation
  11. Results & Discussion
  12. Generative Results
  13. LV Segmentation
  14. Conclusion
  15. Acknowledgements

Figures (1)

  • Figure 1: Synthetic TEE Generation and Image Segmentation Pipeline Using landmarks, desired TEE planes are extracted from each heart model. These ideal slices are the ground truth labels for the synthetic images. Next, pseudo-images are made by adding the acquisition cone and some transformations e.g. Gaussian blurring, shadowing etc., The image synthesis phase concludes with pseudo-images being passed through a trained generator. The synthetic images and their masks can then be used to augment a real dataset for the chosen task