Imposing boundary‐aware prior into CNNs‐based medical image segmentation

Abstract

While convolutional neural networks (CNNs) have become the first choice for the medical image segmentation, they still lack the critical ingredient of incorporating priors, such as smoothness and boundary shapes. The authors tackle the limitation by developing a novel prior that is boundary-aware in two ways: promoting smoothness without blurring object boundaries and punishing prediction errors according to boundary shapes. They bring the boundary-aware property into effect by weighting the prediction gradients and errors with the distance map. Their prior differs from previous approaches that either over-smooth boundaries or tend to produce rough boundaries. They evaluate their prior alongside the cross-entropy (CE) on a cardiac MRI dataset. Compared to CE alone, their prior improves the Dice score by 1.5% and Hausdorff distance by 53%. It also yielded a faster and more stable learning process.

RefineNet‐based 2D and 3D automatic segmentations for clinical target volume and organs at risks for patients with cervical cancer in postoperative radiotherapy

“…Typically, the final feature map output ends up being 32 times smaller in each spatial dimension than the original image. Figure 1b–d demonstrates the encoder–decoder architectures of fully convolutional networks (FCN), U‐Net, and context encoder network (CE‐Net) for comparison 2 4–26 …”

“…As for cervical cancer, three paralleled convolutional neural networks (CNNs) with the same architecture trained following different image preprocessing methods had been applied 17,18 . However, CNNs suffer from the problem of reducing the resolution of original images while increasing the ambiguity of object boundaries inevitably 19 . Recently, the lightweight RefineNet was introduced to refine object detectors for autonomous driving, which generates high‐resolution semantic feature by fusing coarse high‐level features with finer grained low‐level features 20 .…”

“…17,18 However, CNNs suffer from the problem of reducing the resolution of original images while increasing the ambiguity of object boundaries inevitably. 19 Recently, the lightweight RefineNet was introduced to refine object detectors for autonomous driving, which generates high-resolution semantic feature by fusing coarse high-level features with finer grained low-level features. 20 The purpose of this study is to modify the RefineNet and develop a RefineNetPlus3D for the automatic segmentation of CTV and OARs for postoperative cervical cancer based on computed tomography (CT) images, as well as to investigate the accuracy of the RefineNetPlus3D-based automatic segmentation algorithm by comparing it with several other deep learning methods.…”