Unsupervised Cardiac Segmentation Utilizing Synthesized Images from Anatomical Labels

TL;DR

This paper introduces an unsupervised cardiac segmentation method that combines intensity and shape constraints, using synthetic images from anatomical labels to improve segmentation accuracy on cardiac MRI data.

Contribution

The work presents a novel unsupervised framework that integrates intensity and shape constraints, including synthetic image generation from labels, for multi-class cardiac segmentation.

Findings

Achieved Dice scores of 0.5737, 0.7796, and 0.6287 on Myo, LV, and RV.

Demonstrated promising results on MICCAI2019 MSCMR Challenge datasets.

Enhanced segmentation performance without manual annotations.

Abstract

Cardiac segmentation is in great demand for clinical practice. Due to the enormous labor of manual delineation, unsupervised segmentation is desired. The ill-posed optimization problem of this task is inherently challenging, requiring well-designed constraints. In this work, we propose an unsupervised framework for multi-class segmentation with both intensity and shape constraints. Firstly, we extend a conventional non-convex energy function as an intensity constraint and implement it with U-Net. For shape constraint, synthetic images are generated from anatomical labels via image-to-image translation, as shape supervision for the segmentation network. Moreover, augmentation invariance is applied to facilitate the segmentation network to learn the latent features in terms of shape. We evaluated the proposed framework using the public datasets from MICCAI2019 MSCMR Challenge and achieved promising results on cardiac MRIs with Dice scores of 0.5737, 0.7796, and 0.6287 in Myo, LV, and RV, respectively.