Pairwise Semantic Segmentation via Conjugate Fully Convolutional Network

Wang, Renzhen; Cao, Shilei; Ma, Kai; Meng, Deyu; Zheng, Yefeng

doi:10.1007/978-3-030-32226-7_18

Cited by 8 publications

(22 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By essence, the CNN is built to receive inputs of 2D nature, with multiple channels, however, the key difference lies in its output, where the segmentation mask is generated for the center slice only, and the neighboring slices serve only as context and 3D spatial information providers for the model. The idea in itself is not new, it is clearly stated in [40], but many works opt for this method as it harnesses the benefits of both inputs' dimensions and disposes their disadvantages [41][42][43][44][45][46][47], and others.…”

Section: 5d Inputmentioning

confidence: 99%

“…In [58], a complementary network (CompNet) is employed for the segmentation task by attempting to incorporate non-TOI pixels into learning of TOIs ones [99]. A pairwise segmentation technique for sharing supervised segmentation between two paths is investigated by the conjugate FCN (CoFCN) [45], where it takes 2.5D input and learns from adjacent slices explicitly what the segmentation mask should be. In [100], 2D deep belief networks (DBN) is deployed to segment the liver, aided by ASM for post-processing refinement.…”

Section: D Fcnmentioning

confidence: 99%

See 1 more Smart Citation

Automated liver tissues delineation based on machine learning techniques: A survey, current trends and future orientations

Al‐Kababji¹,

Bensaali²,

Dakua³

et al. 2021

Preprint

View full text Add to dashboard Cite

There is no denying how machine learning and computer vision have grown in the recent years. Their highest advantages lie within their automation, suitability, and ability to generate astounding results in a matter of seconds in a reproducible manner. This is aided by the ubiquitous advancements reached in the computing capabilities of current graphical processing units and the highly efficient implementation of such techniques. Hence, in this paper, we survey the key studies that are published between 2014 and 2020, showcasing the different machine learning algorithms researchers have used to segment the liver, hepatic-tumors, and hepatic-vasculature structures. We divide the surveyed studies based on the tissue of interest (hepatic-parenchyma, hepatic-tumors, or hepatic-vessels), highlighting the studies that tackle more than one task simultaneously. Additionally, the machine learning algorithms are classified as either supervised or unsupervised, and further partitioned if the amount of works that fall under a certain scheme is significant. Moreover, different datasets and challenges found in literature and websites, containing masks of the aforementioned tissues, are thoroughly discussed, highlighting the organizers' original contributions, and those of other researchers. Also, the metrics that are used excessively in literature are mentioned in our review stressing their relevancy to the task at hand. Finally, critical challenges and future directions are emphasized for innovative researchers to tackle, exposing gaps that need addressing such as the scarcity of many studies on the vessels' segmentation challenge, and why their absence needs to be dealt with in an accelerated manner.

show abstract

Section: 5d Inputmentioning

confidence: 99%

Section: D Fcnmentioning

confidence: 99%

Automated liver tissues delineation based on machine learning techniques: A survey, current trends and future orientations

Al‐Kababji¹,

Bensaali²,

Dakua³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…With the advent of deep learning, fully convolutional networks (FCNs) have quickly become dominant. These include 2D [36,2], 2.5D [13,43], 3D [28,20,49,46], and hybrid [25,47] FCN-like architectures. Some reported results show that 3D models can improve over 2D ones, but these improvements are sometimes marginal [20].…”

Section: Related Workmentioning

confidence: 99%

“…Compared to improvements in mean DSCs, these worst-case reductions, with commensurate boosts in reliability, can often be more impactful for clinical applications. Unlike CHASe, most prior work on pathological liver segmentation is fully-supervised [2,13,36,25,20,43,46]. For instance, Wang et al report 96.4% DSC on 26 LiTS volumes and Yang et al [46] report 95% DSC on 50 test volumes with unclear healthy vs pathological status.…”

Section: Pathological Liver Segmentationmentioning

confidence: 99%

Co-Heterogeneous and Adaptive Segmentation from Multi-Source and Multi-Phase CT Imaging Data: A Study on Pathological Liver and Lesion Segmentation

Raju¹,

Cheng²,

Yunakai³

et al. 2020

Preprint

View full text Add to dashboard Cite

In medical imaging, organ/pathology segmentation models trained on current publicly available and fully-annotated datasets usually do not well-represent the heterogeneous modalities, phases, pathologies, and clinical scenarios encountered in real environments. On the other hand, there are tremendous amounts of unlabelled patient imaging scans stored by many modern clinical centers. In this work, we present a novel segmentation strategy, co-heterogenous and adaptive segmentation (CHASe), which only requires a small labeled cohort of single phase imaging data to adapt to any unlabeled cohort of heterogenous multi-phase data with possibly new clinical scenarios and pathologies. To do this, we propose a versatile framework that fuses appearance-based semi-supervision, maskbased adversarial domain adaptation, and pseudo-labeling. We also introduce coheterogeneous training, which is a novel integration of co-training and heteromodality learning. We have evaluated CHASe using a clinically comprehensive and challenging dataset of multi-phase computed tomography (CT) imaging studies (1147 patients and 4577 3D volumes). Compared to previous state-of-the-art baselines, CHASe can further improve pathological liver mask Dice-Sørensen coefficients by ranges of 4.2% ∼ 9.4%, depending on the phase combinations: e.g., from 84.6% to 94.0% on non-contrast CTs.

show abstract

“…So far, we are aware of only one work [9] closely related to ours, where the shape priors and inter-subject similarity were both leveraged within a single DL framework.…”

Section: Introductionmentioning

confidence: 99%

Learning Shape Priors by Pairwise Comparison for Robust Semantic Segmentation

Xie

Liu

Cao

et al. 2021

2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI)

Self Cite

View full text Add to dashboard Cite

Semantic segmentation is important in medical image analysis. Inspired by the strong ability of traditional image analysis techniques in capturing shape priors and inter-subject similarity, many deep learning (DL) models have been recently proposed to exploit such prior information and achieved robust performance. However, these two types of important prior information are usually studied separately in existing models. In this paper, we propose a novel DL model to model both type of priors within a single framework. Specifically, we introduce an extra encoder into the classic encoder-decoder structure to form a Siamese structure for the encoders, where one of them takes a target image as input (the image-encoder), and the other concatenates a template image and its foreground regions as input (the templateencoder). The template-encoder encodes the shape priors and appearance characteristics of each foreground class in the template image. A cosine similarity based attention module is proposed to fuse the information from both encoders, to utilize both types of prior information encoded by the template-encoder and model the inter-subject similarity for each foreground class. Extensive experiments on two public datasets demonstrate that our proposed method can produce superior performance to competing methods.

show abstract

Pairwise Semantic Segmentation via Conjugate Fully Convolutional Network

Cited by 8 publications

References 12 publications

Automated liver tissues delineation based on machine learning techniques: A survey, current trends and future orientations

Automated liver tissues delineation based on machine learning techniques: A survey, current trends and future orientations

Co-Heterogeneous and Adaptive Segmentation from Multi-Source and Multi-Phase CT Imaging Data: A Study on Pathological Liver and Lesion Segmentation

Learning Shape Priors by Pairwise Comparison for Robust Semantic Segmentation

Contact Info

Product

Resources

About