Automated classification of benign and malignant cells from lung cytological images using deep convolutional neural network

Atsushi Teramoto, Ayumi Yamada, Yuka Kiriyama, Tetsuya Tsukamoto, Ke Yan, Ling Zhang, Kazuyoshi Imaizumi, Kuniaki Saito, Hiroshi Fujita

研究成果: Article

抄録

Background: Lung cancer is a leading cause of death worldwide, and its early detection is usually performed with low-dose computed tomography. For lesions suspected of abnormality by CT examination, a cytological diagnosis of lung cells collected by biopsy is first performed. However, atypical cells are the major challenge in malignant lung cell classification for cytotechnologists and cytopathologists. In this study, we aimed to automatize the classification of malignant lung cells from microscopic images using a deep convolutional neural network (DCNN). Method: Cytological specimens were prepared with a liquid-based cytology system and stained using the Papanicolaou technique. Images were acquired with a digital still camera attached to a microscope with a 40× objective lens. The original microscopic images were first cropped to obtain image patches with resolution of 224 × 224 pixels. We obtained 306 benign and 315 malignant image patches. To avoid overfitting, 60,000 patch images were generated using data augmentation by applying rotation, flipping, filtering, and color adjustment. DCNN classification was conducted based on a fine-tuned VGG-16 model. We performed patch-based segmentation of malignant regions in the images and evaluated classification performance using threefold cross-validation. Results: The classification sensitivity and specificity were 89.3 and 83.3%, respectively, reaching a performance comparable to that of a cytopathologist. Using the gradient-weighted class activation mapping, we visualized the DCNN identification performance while the network searched for typical benign and malignant cells in images for classification. Conclusions: The proposed method can be useful for accurate and automatic classification of lung cells from pulmonary cytological images.

元の言語English
記事番号100205
ジャーナルInformatics in Medicine Unlocked
16
DOI
出版物ステータスPublished - 01-01-2019

Fingerprint

Lung
Lenses
Cell Biology
Cause of Death
Lung Neoplasms
Color
Tomography
Biopsy
Sensitivity and Specificity

All Science Journal Classification (ASJC) codes

  • Health Informatics

これを引用

Teramoto, Atsushi ; Yamada, Ayumi ; Kiriyama, Yuka ; Tsukamoto, Tetsuya ; Yan, Ke ; Zhang, Ling ; Imaizumi, Kazuyoshi ; Saito, Kuniaki ; Fujita, Hiroshi. / Automated classification of benign and malignant cells from lung cytological images using deep convolutional neural network. :: Informatics in Medicine Unlocked. 2019 ; 巻 16.
@article{ea1fd4505e4f42b3a226f347b6fa38b9,
title = "Automated classification of benign and malignant cells from lung cytological images using deep convolutional neural network",
abstract = "Background: Lung cancer is a leading cause of death worldwide, and its early detection is usually performed with low-dose computed tomography. For lesions suspected of abnormality by CT examination, a cytological diagnosis of lung cells collected by biopsy is first performed. However, atypical cells are the major challenge in malignant lung cell classification for cytotechnologists and cytopathologists. In this study, we aimed to automatize the classification of malignant lung cells from microscopic images using a deep convolutional neural network (DCNN). Method: Cytological specimens were prepared with a liquid-based cytology system and stained using the Papanicolaou technique. Images were acquired with a digital still camera attached to a microscope with a 40× objective lens. The original microscopic images were first cropped to obtain image patches with resolution of 224 × 224 pixels. We obtained 306 benign and 315 malignant image patches. To avoid overfitting, 60,000 patch images were generated using data augmentation by applying rotation, flipping, filtering, and color adjustment. DCNN classification was conducted based on a fine-tuned VGG-16 model. We performed patch-based segmentation of malignant regions in the images and evaluated classification performance using threefold cross-validation. Results: The classification sensitivity and specificity were 89.3 and 83.3{\%}, respectively, reaching a performance comparable to that of a cytopathologist. Using the gradient-weighted class activation mapping, we visualized the DCNN identification performance while the network searched for typical benign and malignant cells in images for classification. Conclusions: The proposed method can be useful for accurate and automatic classification of lung cells from pulmonary cytological images.",
author = "Atsushi Teramoto and Ayumi Yamada and Yuka Kiriyama and Tetsuya Tsukamoto and Ke Yan and Ling Zhang and Kazuyoshi Imaizumi and Kuniaki Saito and Hiroshi Fujita",
year = "2019",
month = "1",
day = "1",
doi = "10.1016/j.imu.2019.100205",
language = "English",
volume = "16",
journal = "Informatics in Medicine Unlocked",
issn = "2352-9148",
publisher = "Elsevier Limited",

}

Automated classification of benign and malignant cells from lung cytological images using deep convolutional neural network. / Teramoto, Atsushi; Yamada, Ayumi; Kiriyama, Yuka; Tsukamoto, Tetsuya; Yan, Ke; Zhang, Ling; Imaizumi, Kazuyoshi; Saito, Kuniaki; Fujita, Hiroshi.

:: Informatics in Medicine Unlocked, 巻 16, 100205, 01.01.2019.

研究成果: Article

TY - JOUR

T1 - Automated classification of benign and malignant cells from lung cytological images using deep convolutional neural network

AU - Teramoto, Atsushi

AU - Yamada, Ayumi

AU - Kiriyama, Yuka

AU - Tsukamoto, Tetsuya

AU - Yan, Ke

AU - Zhang, Ling

AU - Imaizumi, Kazuyoshi

AU - Saito, Kuniaki

AU - Fujita, Hiroshi

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Background: Lung cancer is a leading cause of death worldwide, and its early detection is usually performed with low-dose computed tomography. For lesions suspected of abnormality by CT examination, a cytological diagnosis of lung cells collected by biopsy is first performed. However, atypical cells are the major challenge in malignant lung cell classification for cytotechnologists and cytopathologists. In this study, we aimed to automatize the classification of malignant lung cells from microscopic images using a deep convolutional neural network (DCNN). Method: Cytological specimens were prepared with a liquid-based cytology system and stained using the Papanicolaou technique. Images were acquired with a digital still camera attached to a microscope with a 40× objective lens. The original microscopic images were first cropped to obtain image patches with resolution of 224 × 224 pixels. We obtained 306 benign and 315 malignant image patches. To avoid overfitting, 60,000 patch images were generated using data augmentation by applying rotation, flipping, filtering, and color adjustment. DCNN classification was conducted based on a fine-tuned VGG-16 model. We performed patch-based segmentation of malignant regions in the images and evaluated classification performance using threefold cross-validation. Results: The classification sensitivity and specificity were 89.3 and 83.3%, respectively, reaching a performance comparable to that of a cytopathologist. Using the gradient-weighted class activation mapping, we visualized the DCNN identification performance while the network searched for typical benign and malignant cells in images for classification. Conclusions: The proposed method can be useful for accurate and automatic classification of lung cells from pulmonary cytological images.

AB - Background: Lung cancer is a leading cause of death worldwide, and its early detection is usually performed with low-dose computed tomography. For lesions suspected of abnormality by CT examination, a cytological diagnosis of lung cells collected by biopsy is first performed. However, atypical cells are the major challenge in malignant lung cell classification for cytotechnologists and cytopathologists. In this study, we aimed to automatize the classification of malignant lung cells from microscopic images using a deep convolutional neural network (DCNN). Method: Cytological specimens were prepared with a liquid-based cytology system and stained using the Papanicolaou technique. Images were acquired with a digital still camera attached to a microscope with a 40× objective lens. The original microscopic images were first cropped to obtain image patches with resolution of 224 × 224 pixels. We obtained 306 benign and 315 malignant image patches. To avoid overfitting, 60,000 patch images were generated using data augmentation by applying rotation, flipping, filtering, and color adjustment. DCNN classification was conducted based on a fine-tuned VGG-16 model. We performed patch-based segmentation of malignant regions in the images and evaluated classification performance using threefold cross-validation. Results: The classification sensitivity and specificity were 89.3 and 83.3%, respectively, reaching a performance comparable to that of a cytopathologist. Using the gradient-weighted class activation mapping, we visualized the DCNN identification performance while the network searched for typical benign and malignant cells in images for classification. Conclusions: The proposed method can be useful for accurate and automatic classification of lung cells from pulmonary cytological images.

UR - http://www.scopus.com/inward/record.url?scp=85068530812&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068530812&partnerID=8YFLogxK

U2 - 10.1016/j.imu.2019.100205

DO - 10.1016/j.imu.2019.100205

M3 - Article

AN - SCOPUS:85068530812

VL - 16

JO - Informatics in Medicine Unlocked

JF - Informatics in Medicine Unlocked

SN - 2352-9148

M1 - 100205

ER -