Liver Segmentation on a Variety of Computed Tomography (CT) Images Based on Convolutional Neural Networks Combined with Connected Components

Liver segmentation is relevant for several clinical applications. Automatic liver segmentation using convolutional neural networks (CNNs) has been recently investigated. In this paper, we propose a new approach of combining a largest connected component (LCC) algorithm, as a post-processing step, wi...

Full description

Saved in:
Bibliographic Details
Main Authors: Hoang, Hong Son, Pham, Cam Phuong, Theo, van Walsum, Luu, Manh Ha
Format: Article
Language:English
Published: H. : ĐHQGHN 2020
Subjects:
Online Access:http://repository.vnu.edu.vn/handle/VNU_123/89099
https://doi.org/10.25073/2588-1086/vnucsce.241
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Vietnam National University, Hanoi
Language: English
id oai:112.137.131.14:VNU_123-89099
record_format dspace
spelling oai:112.137.131.14:VNU_123-890992020-06-23T03:46:53Z Liver Segmentation on a Variety of Computed Tomography (CT) Images Based on Convolutional Neural Networks Combined with Connected Components Hoang, Hong Son Pham, Cam Phuong Theo, van Walsum Luu, Manh Ha Liver segmentations CNNs Connected Components Post processing Liver segmentation is relevant for several clinical applications. Automatic liver segmentation using convolutional neural networks (CNNs) has been recently investigated. In this paper, we propose a new approach of combining a largest connected component (LCC) algorithm, as a post-processing step, with CNN approaches to improve liver segmentation accuracy. Specifically, in this study, the algorithm is combined with three well-known CNNs for liver segmentation: FCN-CRF, DRIU and V-net. We perform the experiment on a variety of liver CT images, ranging from non-contrast enhanced CT images to low-dose contrast enhanced CT images. The methods are evaluated using Dice score, Haudorff distance, mean surface distance, and false positive rate between the liver segmentation and the ground truth. The quantitative results demonstrate that the LCC algorithm statistically significantly improves results of the liver segmentation on non-contrast enhanced and low-dose images for all three CNNs. The combination with V-net shows the best performance in Dice score (higher than 90%), while the DRIU network achieves the smallest computation time (2 to 6 seconds) for a single segmentation on average. 2020-06-23T03:46:53Z 2020-06-23T03:46:53Z 2020 Article Hoang, H. S., et al. (2020). Liver Segmentation on a Variety of Computed Tomography (CT) Images Based on Convolutional Neural Networks Combined with Connected Components. VNU Journal of Science: Comp. Science & Com. Eng, Vol. 36, No. 1 (2020) 25-37. 2588-1086 http://repository.vnu.edu.vn/handle/VNU_123/89099 https://doi.org/10.25073/2588-1086/vnucsce.241 en Computer Science and Communication Engineering; application/pdf H. : ĐHQGHN
institution Vietnam National University, Hanoi
building VNU Library & Information Center
country Vietnam
collection VNU Digital Repository
language English
topic Liver segmentations
CNNs
Connected Components
Post processing
spellingShingle Liver segmentations
CNNs
Connected Components
Post processing
Hoang, Hong Son
Pham, Cam Phuong
Theo, van Walsum
Luu, Manh Ha
Liver Segmentation on a Variety of Computed Tomography (CT) Images Based on Convolutional Neural Networks Combined with Connected Components
description Liver segmentation is relevant for several clinical applications. Automatic liver segmentation using convolutional neural networks (CNNs) has been recently investigated. In this paper, we propose a new approach of combining a largest connected component (LCC) algorithm, as a post-processing step, with CNN approaches to improve liver segmentation accuracy. Specifically, in this study, the algorithm is combined with three well-known CNNs for liver segmentation: FCN-CRF, DRIU and V-net. We perform the experiment on a variety of liver CT images, ranging from non-contrast enhanced CT images to low-dose contrast enhanced CT images. The methods are evaluated using Dice score, Haudorff distance, mean surface distance, and false positive rate between the liver segmentation and the ground truth. The quantitative results demonstrate that the LCC algorithm statistically significantly improves results of the liver segmentation on non-contrast enhanced and low-dose images for all three CNNs. The combination with V-net shows the best performance in Dice score (higher than 90%), while the DRIU network achieves the smallest computation time (2 to 6 seconds) for a single segmentation on average.
format Article
author Hoang, Hong Son
Pham, Cam Phuong
Theo, van Walsum
Luu, Manh Ha
author_facet Hoang, Hong Son
Pham, Cam Phuong
Theo, van Walsum
Luu, Manh Ha
author_sort Hoang, Hong Son
title Liver Segmentation on a Variety of Computed Tomography (CT) Images Based on Convolutional Neural Networks Combined with Connected Components
title_short Liver Segmentation on a Variety of Computed Tomography (CT) Images Based on Convolutional Neural Networks Combined with Connected Components
title_full Liver Segmentation on a Variety of Computed Tomography (CT) Images Based on Convolutional Neural Networks Combined with Connected Components
title_fullStr Liver Segmentation on a Variety of Computed Tomography (CT) Images Based on Convolutional Neural Networks Combined with Connected Components
title_full_unstemmed Liver Segmentation on a Variety of Computed Tomography (CT) Images Based on Convolutional Neural Networks Combined with Connected Components
title_sort liver segmentation on a variety of computed tomography (ct) images based on convolutional neural networks combined with connected components
publisher H. : ĐHQGHN
publishDate 2020
url http://repository.vnu.edu.vn/handle/VNU_123/89099
https://doi.org/10.25073/2588-1086/vnucsce.241
_version_ 1680965905324965888