Multi-scale gradual integration CNN for false positive reduction in pulmonary nodule detection

Bum Chae Kim, Jee Seok Yoon, Jun Sik Choi, Heung-Il Suk

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Lung cancer is a global and dangerous disease, and its early detection is crucial for reducing the risks of mortality. In this regard, it has been of great interest in developing a computer-aided system for pulmonary nodules detection as early as possible on thoracic CT scans. In general, a nodule detection system involves two steps: (i) candidate nodule detection at a high sensitivity, which captures many false positives and (ii) false positive reduction from candidates. However, due to the high variation of nodule morphological characteristics and the possibility of mistaking them for neighboring organs, candidate nodule detection remains a challenge. In this study, we propose a novel Multi-scale Gradual Integration Convolutional Neural Network (MGI-CNN), designed with three main strategies: (1) to use multi-scale inputs with different levels of contextual information, (2) to use abstract information inherent in different input scales with gradual integration, and (3) to learn multi-stream feature integration in an end-to-end manner. To verify the efficacy of the proposed network, we conducted exhaustive experiments on the LUNA16 challenge datasets by comparing the performance of the proposed method with state-of-the-art methods in the literature. On two candidate subsets of the LUNA16 dataset, i.e., V1 and V2, our method achieved an average CPM of 0.908 (V1) and 0.942 (V2), outperforming comparable methods by a large margin. Our MGI-CNN is implemented in Python using TensorFlow and the source code is available from https://github.com/ku-milab/MGICNN.

Original languageEnglish
Pages (from-to)1-10
Number of pages10
JournalNeural Networks
Volume115
DOIs
Publication statusPublished - 2019 Jul 1

Fingerprint

Lung
Boidae
Neural networks
Computerized tomography
Computer Systems
Early Diagnosis
Lung Neoplasms
Thorax
Mortality
Experiments
Datasets

Keywords

  • False positive reduction
  • Lung cancer screening
  • Multi-scale convolutional neural network
  • Multi-stream feature integration
  • Pulmonary nodule detection

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Artificial Intelligence

Cite this

Multi-scale gradual integration CNN for false positive reduction in pulmonary nodule detection. / Kim, Bum Chae; Yoon, Jee Seok; Choi, Jun Sik; Suk, Heung-Il.

In: Neural Networks, Vol. 115, 01.07.2019, p. 1-10.

Research output: Contribution to journalArticle

Kim, Bum Chae ; Yoon, Jee Seok ; Choi, Jun Sik ; Suk, Heung-Il. / Multi-scale gradual integration CNN for false positive reduction in pulmonary nodule detection. In: Neural Networks. 2019 ; Vol. 115. pp. 1-10.
@article{c8954044dc7b43d7948d16a293ab9769,
title = "Multi-scale gradual integration CNN for false positive reduction in pulmonary nodule detection",
abstract = "Lung cancer is a global and dangerous disease, and its early detection is crucial for reducing the risks of mortality. In this regard, it has been of great interest in developing a computer-aided system for pulmonary nodules detection as early as possible on thoracic CT scans. In general, a nodule detection system involves two steps: (i) candidate nodule detection at a high sensitivity, which captures many false positives and (ii) false positive reduction from candidates. However, due to the high variation of nodule morphological characteristics and the possibility of mistaking them for neighboring organs, candidate nodule detection remains a challenge. In this study, we propose a novel Multi-scale Gradual Integration Convolutional Neural Network (MGI-CNN), designed with three main strategies: (1) to use multi-scale inputs with different levels of contextual information, (2) to use abstract information inherent in different input scales with gradual integration, and (3) to learn multi-stream feature integration in an end-to-end manner. To verify the efficacy of the proposed network, we conducted exhaustive experiments on the LUNA16 challenge datasets by comparing the performance of the proposed method with state-of-the-art methods in the literature. On two candidate subsets of the LUNA16 dataset, i.e., V1 and V2, our method achieved an average CPM of 0.908 (V1) and 0.942 (V2), outperforming comparable methods by a large margin. Our MGI-CNN is implemented in Python using TensorFlow and the source code is available from https://github.com/ku-milab/MGICNN.",
keywords = "False positive reduction, Lung cancer screening, Multi-scale convolutional neural network, Multi-stream feature integration, Pulmonary nodule detection",
author = "Kim, {Bum Chae} and Yoon, {Jee Seok} and Choi, {Jun Sik} and Heung-Il Suk",
year = "2019",
month = "7",
day = "1",
doi = "10.1016/j.neunet.2019.03.003",
language = "English",
volume = "115",
pages = "1--10",
journal = "Neural Networks",
issn = "0893-6080",
publisher = "Elsevier Limited",

}

TY - JOUR

T1 - Multi-scale gradual integration CNN for false positive reduction in pulmonary nodule detection

AU - Kim, Bum Chae

AU - Yoon, Jee Seok

AU - Choi, Jun Sik

AU - Suk, Heung-Il

PY - 2019/7/1

Y1 - 2019/7/1

N2 - Lung cancer is a global and dangerous disease, and its early detection is crucial for reducing the risks of mortality. In this regard, it has been of great interest in developing a computer-aided system for pulmonary nodules detection as early as possible on thoracic CT scans. In general, a nodule detection system involves two steps: (i) candidate nodule detection at a high sensitivity, which captures many false positives and (ii) false positive reduction from candidates. However, due to the high variation of nodule morphological characteristics and the possibility of mistaking them for neighboring organs, candidate nodule detection remains a challenge. In this study, we propose a novel Multi-scale Gradual Integration Convolutional Neural Network (MGI-CNN), designed with three main strategies: (1) to use multi-scale inputs with different levels of contextual information, (2) to use abstract information inherent in different input scales with gradual integration, and (3) to learn multi-stream feature integration in an end-to-end manner. To verify the efficacy of the proposed network, we conducted exhaustive experiments on the LUNA16 challenge datasets by comparing the performance of the proposed method with state-of-the-art methods in the literature. On two candidate subsets of the LUNA16 dataset, i.e., V1 and V2, our method achieved an average CPM of 0.908 (V1) and 0.942 (V2), outperforming comparable methods by a large margin. Our MGI-CNN is implemented in Python using TensorFlow and the source code is available from https://github.com/ku-milab/MGICNN.

AB - Lung cancer is a global and dangerous disease, and its early detection is crucial for reducing the risks of mortality. In this regard, it has been of great interest in developing a computer-aided system for pulmonary nodules detection as early as possible on thoracic CT scans. In general, a nodule detection system involves two steps: (i) candidate nodule detection at a high sensitivity, which captures many false positives and (ii) false positive reduction from candidates. However, due to the high variation of nodule morphological characteristics and the possibility of mistaking them for neighboring organs, candidate nodule detection remains a challenge. In this study, we propose a novel Multi-scale Gradual Integration Convolutional Neural Network (MGI-CNN), designed with three main strategies: (1) to use multi-scale inputs with different levels of contextual information, (2) to use abstract information inherent in different input scales with gradual integration, and (3) to learn multi-stream feature integration in an end-to-end manner. To verify the efficacy of the proposed network, we conducted exhaustive experiments on the LUNA16 challenge datasets by comparing the performance of the proposed method with state-of-the-art methods in the literature. On two candidate subsets of the LUNA16 dataset, i.e., V1 and V2, our method achieved an average CPM of 0.908 (V1) and 0.942 (V2), outperforming comparable methods by a large margin. Our MGI-CNN is implemented in Python using TensorFlow and the source code is available from https://github.com/ku-milab/MGICNN.

KW - False positive reduction

KW - Lung cancer screening

KW - Multi-scale convolutional neural network

KW - Multi-stream feature integration

KW - Pulmonary nodule detection

UR - http://www.scopus.com/inward/record.url?scp=85063187357&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063187357&partnerID=8YFLogxK

U2 - 10.1016/j.neunet.2019.03.003

DO - 10.1016/j.neunet.2019.03.003

M3 - Article

C2 - 30909118

AN - SCOPUS:85063187357

VL - 115

SP - 1

EP - 10

JO - Neural Networks

JF - Neural Networks

SN - 0893-6080

ER -