Cardiac Segmentation on CT Images through Shape-Aware Contour Attentions

TL;DR

The proposed shape-aware contour attention mechanism demonstrates that distance transformation and boundary features improve the actual attention map to strengthen the responses in the boundary area.

Abstract

Cardiac segmentation of atriums, ventricles, and myocardium in computed tomography (CT) images is an important first-line task for presymptomatic cardiovascular disease diagnosis. In several recent studies, deep learning models have shown significant breakthroughs in medical image segmentation tasks. Unlike other organs such as the lungs and liver, the cardiac organ consists of multiple substructures, i.e., ventricles, atriums, aortas, arteries, veins, and myocardium. These cardiac substructures are proximate to each other and have indiscernible boundaries (i.e., homogeneous intensity values), making it difficult for the segmentation network focus on the boundaries between the substructures. In this paper, to improve the segmentation accuracy between proximate organs, we introduce a novel model to exploit shape and boundary-aware features. We primarily propose a shape-aware attention module, that exploits distance regression, which can guide the model to focus on the edges between substructures so that it can outperform the conventional contour-based attention method. In the experiments, we used the Multi-Modality Whole Heart Segmentation dataset that has 20 CT cardiac images for training and validation, and 40 CT cardiac images for testing. The experimental results show that the proposed network produces more accurate results than state-of-the-art networks by improving the Dice similarity coefficient score by 4.97%. Our proposed shape-aware contour attention mechanism demonstrates that distance transformation and boundary features improve the actual attention map to strengthen the responses in the boundary area. Moreover, our proposed method significantly reduces the false-positive responses of the final output, resulting in accurate segmentation.

Figures (11)

• Figure 1: Cardiac computed tomography image, manual segmentation labels of cardiac substructures, and overlapped images
• Figure 2: Overall architecture of the proposed convolutional neural network model
• Figure 3: Schematic visualization of contour- and distance-transform-guided attention network (CDA-Net). The blue and yellow cubes correspond to feature maps. The dotted line denotes each feature map from the contour transition network and distance transform transition network.
• Figure 4: Diagram of V-transition module introduced in chung2020liver, which is composed of a single-level encoder-decoder architecture and channel-wise attention. The separable conv block consists of a convolutional layer with groups, batch normalization, and a non-linear activation function, which reduces the computation times and parameters. The single-level encoder-decoder architecture supports the parameters to be trained with multi-scale spatial features, and the channel-wise attention refines and compresses the intermediate features.
• Figure 5: Example of the utilization of penalty energy to refine contour and distance transformation (DT) features. White color indicates a higher value, and the red line indicates the predicted contour. The colored blobs represent equal low probabilities but are colored differently to distinguish between multiple organs. (a) DT features, (b) inversion of clamped DT features, and (c) contour features; (d), (e), and (f) indicate the cases that may appear during the training; (g) shows the trend of our penalty energy in terms of the similarity, which indicates the contour loss. The case shown in (d) has a high penalty as the contour features are overlapping the background area, which has high values. Similarly, the case illustrated in (f) has a low penalty energy. The penalty energy (\ref{['eq:penalty']}) performs an important role as the contour feature responses do not appear in the background to identify the exact boundary line.