Structure Boundary Preserving Segmentation for Medical Image With Ambiguous Boundary

Lee, Hong Joo; Kim, Jung Uk; Lee, Sang‐Min; Kim, Hak Gu; Ro, Yong Man

doi:10.1109/cvpr42600.2020.00487

Cited by 104 publications

(58 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…3, we employ the dilated convolution [8] to exploit different receptive fields after the decoder of student model. The fusion of outputs from different receptive fields is regarded as an effective feature extraction operation [9,10]. To balance the training time and performance, we use three dilated convolution networks, namely d i (•) with a dilation rate r i and filter size s i × s i (i = 1, 2, 3).…”

Section: Learn Model Confidence Under the Guidance Of The True Class Probabilitymentioning

confidence: 99%

Semi-Supervised Skin Lesion Segmentation with Learning Model Confidence

Xie

Zheng

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Segmentation of skin lesions is important for disease diagnoses and treatment planning. Over the years, semisupervised methods using pseudo labels have boosted the segmentation performance with limited labeled data and abundant unlabeled data. However, the unreliable targets in pseudo labels might lead to meaningless guidance for unlabeled data. In this paper, to solve this issue, we propose a novel confidence aware semi-supervised learning method based on a mean teacher scheme. Concretely, we design a confidence module to predict the model confidence guided by the True Class Probability. Then in the mean teacher framework, the student model gradually learns trustworthy targets from teacher model. To further improve the segmentation quality, we fine-tune the student model with reliable content in pseudo labels. We conduct extensive experiments on 2018 ISIC skin lesion segmentation dataset and our method outperforms other state-of-the-art semi-supervised approaches.

show abstract

Section: Learn Model Confidence Under the Guidance Of The True Class Probabilitymentioning

confidence: 99%

Semi-Supervised Skin Lesion Segmentation with Learning Model Confidence

Xie

Zheng

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…We expect that expert knowledge about stroke boundary information is embedded in the fully automatic image binarization model. To embed expert knowledge of structure boundary in the model, authors in [34] propose shape boundary-aware evaluator based on the discrimination network in an adversarial way without the user interaction. For this purpose, we propose a discrimination network based on structure stroke boundary information, as shown in Fig.…”

Section: ) Stroke Boundary Information-based Discrimination Networkmentioning

confidence: 99%

Document Image Binarization With Stroke Boundary Feature Guided Network

Dang

Lee

2021

IEEE Access

View full text Add to dashboard Cite

Text is the most crucial element in a document image but is often disconnected in document image binarization. Most of the previous methods based on deep learning do not focus on structure information such as stroke boundary, leading to disconnected strokes when the stroke is ambiguous or weak. In this paper, we propose a multi-task learning with an auxiliary task for learning stroke boundary features in an adversarial manner. The learned boundary features are integrated into the main task for the binarization. Specifically, in the first step, in addition to using shared global location features with the main task, the auxiliary task leverages additional local edges to obtain stroke boundary features. In the second step, we use adversarial loss based on boundary ground truth to supervise the obtained stroke boundary feature in the auxiliary task. The adversarial training is to embed expert knowledge, especially structure information, in the model. In the third step, the learned boundary feature from the auxiliary task supports the main task directly. The fusion module of the main task refines the final binarized image. Experiments show that our method achieves better-preserved stroke and better performance than existing methods on benchmark H-DBCO and DIBCO datasets.

show abstract

“…2(b). Many existing segmentation models [5,7,8,9] design their boundary awareness modules with effective supervised losses. However, it is difficult to achieve satisfactory result due to the extremely unbalanced pixel distribution between boundary pix-els and other pixels, especially for lung nodule segmentation.…”

Section: Deep Feature Decoupling Modulementioning

confidence: 99%

“…To better capture the boundary information in medical images, a new loss is proposed in [7] to learn the boundary information by combining contour length and region information. Lee et al [8] propose a structure boundary preserving method with an adversarial loss to predict the boundary of the medical object. Fan et al [9] add attention mechanisms to their framework to mine the boundary information of the target object.…”

Section: Introductionmentioning

confidence: 99%

DFDM: A Deep Feature Decoupling Module for Lung Nodule Segmentation

Chen

Wang

Huang

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

In this paper, we propose a novel feature decoupling method to tackle two critical problems in the lung nodule segmentation task: (i) ambiguity of nodule boundary leads to the imprecise segmentation boundary and (ii) the high false positive rate of segmentation result. Our motivation is that an accurate segmentation network needs explicitly modeling the nodule boundary and texture information, and suppressing the noise information. To do so, a novel Deep Feature Decoupling Module (DFDM) is proposed to decouple the nodule boundary, noise, and texture information from the original feature maps. The decoupled boundary and texture information is used to benefit the segmentation, and the noise information is removed from the input features to reduce the false positive rate. The proposed DFDM consists of three parallel branches, including Boundary Sensitive Branch (BSB), Noise Removal Branch (NRB), and Texture Preserving Branch (TPB) to decouple the mentioned three information, respectively. In particular, we design our BSB with a novel architecture to effectively capture the boundary information of lung nodules. We apply the proposed DFDM to the U-Net architecture and achieve convincing segmentation results on the LIDC-IDRI dataset. Code and models are available at https://github.com/chinichenw/DFDM.

show abstract

Structure Boundary Preserving Segmentation for Medical Image With Ambiguous Boundary

Cited by 104 publications

References 27 publications

Semi-Supervised Skin Lesion Segmentation with Learning Model Confidence

Semi-Supervised Skin Lesion Segmentation with Learning Model Confidence

Document Image Binarization With Stroke Boundary Feature Guided Network

DFDM: A Deep Feature Decoupling Module for Lung Nodule Segmentation

Contact Info

Product

Resources

About