Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Automatic Abdominal MRI Organ Segmentation: Leveraging Synthesized Data Generated From CT Labels

Cosmin Ciausu, Deepa Krishnaswamy, Benjamin Billot, Steve Pieper, Ron Kikinis, Andrey Fedorov

TL;DR

A modality-agnostic domain randomization approach is used, utilizing CT label maps to generate synthetic images on-the-fly during training, further used to train a U-Net segmentation network for abdominal organs segmentation.

Abstract

Deep learning has shown great promise in the ability to automatically annotate organs in magnetic resonance imaging (MRI) scans, for example, of the brain. However, despite advancements in the field, the ability to accurately segment abdominal organs remains difficult across MR. In part, this may be explained by the much greater variability in image appearance and severely limited availability of training labels. The inherent nature of computed tomography (CT) scans makes it easier to annotate, resulting in a larger availability of expert annotations for the latter. We leverage a modality-agnostic domain randomization approach, utilizing CT label maps to generate synthetic images on-the-fly during training, further used to train a U-Net segmentation network for abdominal organs segmentation. Our approach shows comparable results compared to fully-supervised segmentation methods trained on MR data. Our method results in Dice scores of 0.90 (0.08) and 0.91 (0.08) for the right and left kidney respectively, compared to a pretrained nnU-Net model yielding 0.87 (0.20) and 0.91 (0.03). We will make our code publicly available.

Towards Automatic Abdominal MRI Organ Segmentation: Leveraging Synthesized Data Generated From CT Labels

TL;DR

A modality-agnostic domain randomization approach is used, utilizing CT label maps to generate synthetic images on-the-fly during training, further used to train a U-Net segmentation network for abdominal organs segmentation.

Abstract

Deep learning has shown great promise in the ability to automatically annotate organs in magnetic resonance imaging (MRI) scans, for example, of the brain. However, despite advancements in the field, the ability to accurately segment abdominal organs remains difficult across MR. In part, this may be explained by the much greater variability in image appearance and severely limited availability of training labels. The inherent nature of computed tomography (CT) scans makes it easier to annotate, resulting in a larger availability of expert annotations for the latter. We leverage a modality-agnostic domain randomization approach, utilizing CT label maps to generate synthetic images on-the-fly during training, further used to train a U-Net segmentation network for abdominal organs segmentation. Our approach shows comparable results compared to fully-supervised segmentation methods trained on MR data. Our method results in Dice scores of 0.90 (0.08) and 0.91 (0.08) for the right and left kidney respectively, compared to a pretrained nnU-Net model yielding 0.87 (0.20) and 0.91 (0.03). We will make our code publicly available.
Paper Structure (14 sections, 2 figures, 4 tables, 1 algorithm)

This paper contains 14 sections, 2 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Methods
  3. Approach
  4. Datasets
  5. Training
  6. Validation and testing
  7. Pre-processing for target labels selection
  8. Pre-processing for synthetic data generation
  9. Training procedure and evaluation
  10. Results and Discussion
  11. Experimental Results
  12. Ablation study results
  13. Comparison to publicly available methods
  14. Conclusion

Figures (2)

  • Figure 1: Flowchart of the method for abdominal segmentation. The model requires as an input training label maps as input which are used to create synthetically generated images and corresponding label maps on the fly. These are used to train a U-Net model using a soft Dice loss. The final segmentation map of the anatomical structures of interests are obtained using the trained model.
  • Figure 2: Qualitative results of the proposed method on a subject from AMOS MR. The left view shows the expert radiologist annotations, and the right one shows the results from our baseline method. Here we have high agreement between many abdominal regions, including the liver in red, spleen in green, kidneys in yellow and brown. Our method segments 26 structures in the abdominal area.