Single-Channel Sparse Non-Negative Blind Source Separation Method for Automatic 3-D Delineation of Lung Tumor in PET Images

Ivica Kopriva; Wei Ju; Bin Zhang; Fei Shi; Dehui Xiang; Kai Yu; Ximing Wang; Ulas Bagci; Xinjian Chen

doi:10.1109/JBHI.2016.2624798

Single-Channel Sparse Non-Negative Blind Source Separation Method for Automatic 3-D Delineation of Lung Tumor in PET Images

Ivica Kopriva, Wei Ju, Bin Zhang, Fei Shi, Dehui Xiang, Kai Yu, Ximing Wang, Ulas Bagci, Xinjian Chen^*

^*Corresponding author for this work

Radiology

Research output: Contribution to journal › Article › peer-review

13 Scopus citations

Abstract

In this paper, we propose a novel method for single-channel blind separation of nonoverlapped sources and, to the best of our knowledge, apply it for the first time to automatic segmentation of lung tumors in positron emission tomography (PET) images. Our approach first converts a 3-D PET image into a pseudo-multichannel image. Afterward, regularization free sparseness constrained non-negative matrix factorization is used to separate tumor from other tissues. By using complexity based criterion, we select tumor component as the one with minimal complexity. We have compared the proposed method with threshold based on 40% and 50% maximum standardized uptake value (SUV), graph cuts (GC), random walks (RW), and affinity propagation (AP) algorithms on 18 nonsmall cell lung cancer datasets with respect to ground truth (GT) provided by two radiologists. Dice similarity coefficient averaged with respect to two GTs is: 0.78 ± 0.12 by the proposed algorithm, 0.78 ± 0.1 by GC, 0.77 ± 0.13 by AP, 0.77 ± 0.07 by RW, and 0.75 ± 0.13 by 50% maximum SUV threshold. Since the proposed method achieved performance comparable with interactive methods, considering the unique challenges of lung tumor segmentation from PET images, our findings support possibility of using our fully automated method in routine clinics. The source codes will be available at www.mipav.net/English/research/research.html.

Original language	English (US)
Article number	7733134
Pages (from-to)	1656-1666
Number of pages	11
Journal	IEEE Journal of Biomedical and Health Informatics
Volume	21
Issue number	6
DOIs	https://doi.org/10.1109/JBHI.2016.2624798
State	Published - Nov 2017

Funding

Manuscript received June 6, 2016; revised September 22, 2016 and October 31, 2016; accepted November 1, 2016. Date of publication November 3, 2016; date of current version November 3, 2017. This work was supported in part by the bilateral Chinese–Croatian Grant “Dealina-tion of lung tumor though nonlinear decomposition of PET/CT image;” in part by the Croatian Science Foundation Grant “Structured Decompositions of Empirical Data for Computationally-Assisted Diagnoses of Disease;” in part by the National Basic Research Program of China (973 Program) under Grant 2014CB748600; and in part by the National Natural Science Foundation of China under Grant 81371629. (Corresponding author: Xinjian Chen.) I. Kopriva is with the Division of Electronics, Ruđer Bosˇković Institute, Zagreb 10000, Croatia (e-mail: [email protected]).

Keywords

Lung tumor delineation
non-negative matrix factorization (NMF)
positron emission tomography (PET)
single-channel blind source separation
sparseness

ASJC Scopus subject areas

Computer Science Applications
Health Informatics
Electrical and Electronic Engineering
Health Information Management

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1109/JBHI.2016.2624798

Cite this

@article{e3a9dddb22c24c6089d6f1d06e367b4d,

title = "Single-Channel Sparse Non-Negative Blind Source Separation Method for Automatic 3-D Delineation of Lung Tumor in PET Images",

abstract = "In this paper, we propose a novel method for single-channel blind separation of nonoverlapped sources and, to the best of our knowledge, apply it for the first time to automatic segmentation of lung tumors in positron emission tomography (PET) images. Our approach first converts a 3-D PET image into a pseudo-multichannel image. Afterward, regularization free sparseness constrained non-negative matrix factorization is used to separate tumor from other tissues. By using complexity based criterion, we select tumor component as the one with minimal complexity. We have compared the proposed method with threshold based on 40% and 50% maximum standardized uptake value (SUV), graph cuts (GC), random walks (RW), and affinity propagation (AP) algorithms on 18 nonsmall cell lung cancer datasets with respect to ground truth (GT) provided by two radiologists. Dice similarity coefficient averaged with respect to two GTs is: 0.78 ± 0.12 by the proposed algorithm, 0.78 ± 0.1 by GC, 0.77 ± 0.13 by AP, 0.77 ± 0.07 by RW, and 0.75 ± 0.13 by 50% maximum SUV threshold. Since the proposed method achieved performance comparable with interactive methods, considering the unique challenges of lung tumor segmentation from PET images, our findings support possibility of using our fully automated method in routine clinics. The source codes will be available at www.mipav.net/English/research/research.html.",

keywords = "Lung tumor delineation, non-negative matrix factorization (NMF), positron emission tomography (PET), single-channel blind source separation, sparseness",

author = "Ivica Kopriva and Wei Ju and Bin Zhang and Fei Shi and Dehui Xiang and Kai Yu and Ximing Wang and Ulas Bagci and Xinjian Chen",

note = "Publisher Copyright: {\textcopyright} 2013 IEEE.",

year = "2017",

month = nov,

doi = "10.1109/JBHI.2016.2624798",

language = "English (US)",

volume = "21",

pages = "1656--1666",

journal = "IEEE Journal of Biomedical and Health Informatics",

issn = "2168-2194",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

number = "6",

}

TY - JOUR

T1 - Single-Channel Sparse Non-Negative Blind Source Separation Method for Automatic 3-D Delineation of Lung Tumor in PET Images

AU - Kopriva, Ivica

AU - Ju, Wei

AU - Zhang, Bin

AU - Shi, Fei

AU - Xiang, Dehui

AU - Yu, Kai

AU - Wang, Ximing

AU - Bagci, Ulas

AU - Chen, Xinjian

PY - 2017/11

Y1 - 2017/11

N2 - In this paper, we propose a novel method for single-channel blind separation of nonoverlapped sources and, to the best of our knowledge, apply it for the first time to automatic segmentation of lung tumors in positron emission tomography (PET) images. Our approach first converts a 3-D PET image into a pseudo-multichannel image. Afterward, regularization free sparseness constrained non-negative matrix factorization is used to separate tumor from other tissues. By using complexity based criterion, we select tumor component as the one with minimal complexity. We have compared the proposed method with threshold based on 40% and 50% maximum standardized uptake value (SUV), graph cuts (GC), random walks (RW), and affinity propagation (AP) algorithms on 18 nonsmall cell lung cancer datasets with respect to ground truth (GT) provided by two radiologists. Dice similarity coefficient averaged with respect to two GTs is: 0.78 ± 0.12 by the proposed algorithm, 0.78 ± 0.1 by GC, 0.77 ± 0.13 by AP, 0.77 ± 0.07 by RW, and 0.75 ± 0.13 by 50% maximum SUV threshold. Since the proposed method achieved performance comparable with interactive methods, considering the unique challenges of lung tumor segmentation from PET images, our findings support possibility of using our fully automated method in routine clinics. The source codes will be available at www.mipav.net/English/research/research.html.

AB - In this paper, we propose a novel method for single-channel blind separation of nonoverlapped sources and, to the best of our knowledge, apply it for the first time to automatic segmentation of lung tumors in positron emission tomography (PET) images. Our approach first converts a 3-D PET image into a pseudo-multichannel image. Afterward, regularization free sparseness constrained non-negative matrix factorization is used to separate tumor from other tissues. By using complexity based criterion, we select tumor component as the one with minimal complexity. We have compared the proposed method with threshold based on 40% and 50% maximum standardized uptake value (SUV), graph cuts (GC), random walks (RW), and affinity propagation (AP) algorithms on 18 nonsmall cell lung cancer datasets with respect to ground truth (GT) provided by two radiologists. Dice similarity coefficient averaged with respect to two GTs is: 0.78 ± 0.12 by the proposed algorithm, 0.78 ± 0.1 by GC, 0.77 ± 0.13 by AP, 0.77 ± 0.07 by RW, and 0.75 ± 0.13 by 50% maximum SUV threshold. Since the proposed method achieved performance comparable with interactive methods, considering the unique challenges of lung tumor segmentation from PET images, our findings support possibility of using our fully automated method in routine clinics. The source codes will be available at www.mipav.net/English/research/research.html.

KW - Lung tumor delineation

KW - non-negative matrix factorization (NMF)

KW - positron emission tomography (PET)

KW - single-channel blind source separation

KW - sparseness

UR - http://www.scopus.com/inward/record.url?scp=85035784242&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85035784242&partnerID=8YFLogxK

U2 - 10.1109/JBHI.2016.2624798

DO - 10.1109/JBHI.2016.2624798

M3 - Article

C2 - 27834658

AN - SCOPUS:85035784242

SN - 2168-2194

VL - 21

SP - 1656

EP - 1666

JO - IEEE Journal of Biomedical and Health Informatics

JF - IEEE Journal of Biomedical and Health Informatics

IS - 6

M1 - 7733134

ER -

Single-Channel Sparse Non-Negative Blind Source Separation Method for Automatic 3-D Delineation of Lung Tumor in PET Images

Abstract

Funding

Keywords

ASJC Scopus subject areas

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this