Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

BraSyn 2023 challenge: Missing MRI synthesis and the effect of different learning objectives

Ivo M. Baltruschat, Parvaneh Janbakhshi, Matthias Lenga

TL;DR

This study proposes investigating the effectiveness of a commonly used deep learning framework, such as Pix2Pix, trained under the supervision of different image-quality loss functions, and systematically studies the impact of various loss functions in the multi-sequence MR image synthesis setting of the BraSyn challenge.

Abstract

This work addresses the Brain Magnetic Resonance Image Synthesis for Tumor Segmentation (BraSyn) challenge, which was hosted as part of the Brain Tumor Segmentation (BraTS) challenge in 2023. In this challenge, researchers are invited to synthesize a missing magnetic resonance image sequence, given other available sequences, to facilitate tumor segmentation pipelines trained on complete sets of image sequences. This problem can be tackled using deep learning within the framework of paired image-to-image translation. In this study, we propose investigating the effectiveness of a commonly used deep learning framework, such as Pix2Pix, trained under the supervision of different image-quality loss functions. Our results indicate that the use of different loss functions significantly affects the synthesis quality. We systematically study the impact of various loss functions in the multi-sequence MR image synthesis setting of the BraSyn challenge. Furthermore, we demonstrate how image synthesis performance can be optimized by combining different learning objectives beneficially.

BraSyn 2023 challenge: Missing MRI synthesis and the effect of different learning objectives

TL;DR

This study proposes investigating the effectiveness of a commonly used deep learning framework, such as Pix2Pix, trained under the supervision of different image-quality loss functions, and systematically studies the impact of various loss functions in the multi-sequence MR image synthesis setting of the BraSyn challenge.

Abstract

This work addresses the Brain Magnetic Resonance Image Synthesis for Tumor Segmentation (BraSyn) challenge, which was hosted as part of the Brain Tumor Segmentation (BraTS) challenge in 2023. In this challenge, researchers are invited to synthesize a missing magnetic resonance image sequence, given other available sequences, to facilitate tumor segmentation pipelines trained on complete sets of image sequences. This problem can be tackled using deep learning within the framework of paired image-to-image translation. In this study, we propose investigating the effectiveness of a commonly used deep learning framework, such as Pix2Pix, trained under the supervision of different image-quality loss functions. Our results indicate that the use of different loss functions significantly affects the synthesis quality. We systematically study the impact of various loss functions in the multi-sequence MR image synthesis setting of the BraSyn challenge. Furthermore, we demonstrate how image synthesis performance can be optimized by combining different learning objectives beneficially.
Paper Structure (19 sections, 6 equations, 3 figures, 3 tables)

This paper contains 19 sections, 6 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Dataset
  4. Networks
  5. Synthesis (generator) network
  6. Discriminator
  7. Training
  8. L1 loss
  9. Masked L1 loss
  10. Adversarial loss
  11. SSIM loss
  12. Perceptual VGG loss
  13. Frequency loss
  14. Inference
  15. Results and Discussion
  16. ...and 4 more sections

Figures (3)

  • Figure 1: The used synthesis framework for predicting an exemplary target sequence, e.g., T1-C from input sequences, i.e., T1-N, T2-W, and T2-F.
  • Figure 2: Randomly selected examples from our development set. Each pair (e.g., T1-N and T1-synN) shows the original image and the corresponding synthetic image. All images are histogram normalized and each pair has the same visualization settings.
  • Figure 3: Dice score results for the test set (reported by challenge organizers). Team 1 is our submitted winning solution.