Diagnosis of Thyroid Nodules: Performance of a Deep Learning Convolutional Neural Network Model vs. Radiologists

Vivian Y. Park; Kyunghwa Han; Yeong Kyeong Seong; Moon Ho Park; Eun Kyung Kim; Hee Jung Moon; Jung Hyun Yoon; Jin Young Kwak

doi:10.1038/s41598-019-54434-1

Diagnosis of Thyroid Nodules: Performance of a Deep Learning Convolutional Neural Network Model vs. Radiologists

Vivian Y. Park, Kyunghwa Han, Yeong Kyeong Seong, Moon Ho Park, Eun Kyung Kim, Hee Jung Moon, Jung Hyun Yoon, Jin Young Kwak

Department of Radiology

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

54 Citations (Scopus)

Abstract

Computer-aided diagnosis (CAD) systems hold potential to improve the diagnostic accuracy of thyroid ultrasound (US). We aimed to develop a deep learning-based US CAD system (dCAD) for the diagnosis of thyroid nodules and compare its performance with those of a support vector machine (SVM)-based US CAD system (sCAD) and radiologists. dCAD was developed by using US images of 4919 thyroid nodules from three institutions. Its diagnostic performance was prospectively evaluated between June 2016 and February 2017 in 286 nodules, and was compared with those of sCAD and radiologists, using logistic regression with the generalized estimating equation. Subgroup analyses were performed according to experience level and separately for small thyroid nodules 1–2 cm. There was no difference in overall sensitivity, specificity, positive predictive value (PPV), negative predictive value and accuracy (all p > 0.05) between radiologists and dCAD. Radiologists and dCAD showed higher specificity, PPV, and accuracy than sCAD (all p < 0.001). In small nodules, experienced radiologists showed higher specificity, PPV and accuracy than sCAD (all p < 0.05). In conclusion, dCAD showed overall comparable diagnostic performance with radiologists and assessed thyroid nodules more effectively than sCAD, without loss of sensitivity.

Original language	English
Article number	17843
Journal	Scientific reports
Volume	9
Issue number	1
DOIs	https://doi.org/10.1038/s41598-019-54434-1
Publication status	Published - 2019 Dec 1

Bibliographical note

Funding Information:
This study was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) by the Ministry of Education (2016R1D1A1B03930375) and by the research fund of Samsung Electronics Co., Ltd.

Funding Information:
This prospective study was supported by a grant from Samsung Medison Co. in Seoul, South Korea, which also provided the equipment for this study. The study protocol was reviewed and approved by the Institutional Review Board of Severance hospital. Written informed consent was obtained from all patients before each US examination. All methods were performed in accordance with the relevant guidelines and regulations.

Publisher Copyright:
© 2019, The Author(s).

All Science Journal Classification (ASJC) codes

General

Access to Document

10.1038/s41598-019-54434-1

Cite this

@article{54e4d8214c95403f91e9c73e866d41e1,

title = "Diagnosis of Thyroid Nodules: Performance of a Deep Learning Convolutional Neural Network Model vs. Radiologists",

abstract = "Computer-aided diagnosis (CAD) systems hold potential to improve the diagnostic accuracy of thyroid ultrasound (US). We aimed to develop a deep learning-based US CAD system (dCAD) for the diagnosis of thyroid nodules and compare its performance with those of a support vector machine (SVM)-based US CAD system (sCAD) and radiologists. dCAD was developed by using US images of 4919 thyroid nodules from three institutions. Its diagnostic performance was prospectively evaluated between June 2016 and February 2017 in 286 nodules, and was compared with those of sCAD and radiologists, using logistic regression with the generalized estimating equation. Subgroup analyses were performed according to experience level and separately for small thyroid nodules 1–2 cm. There was no difference in overall sensitivity, specificity, positive predictive value (PPV), negative predictive value and accuracy (all p > 0.05) between radiologists and dCAD. Radiologists and dCAD showed higher specificity, PPV, and accuracy than sCAD (all p < 0.001). In small nodules, experienced radiologists showed higher specificity, PPV and accuracy than sCAD (all p < 0.05). In conclusion, dCAD showed overall comparable diagnostic performance with radiologists and assessed thyroid nodules more effectively than sCAD, without loss of sensitivity.",

author = "Park, {Vivian Y.} and Kyunghwa Han and Seong, {Yeong Kyeong} and Park, {Moon Ho} and Kim, {Eun Kyung} and Moon, {Hee Jung} and Yoon, {Jung Hyun} and Kwak, {Jin Young}",

note = "Funding Information: This study was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) by the Ministry of Education (2016R1D1A1B03930375) and by the research fund of Samsung Electronics Co., Ltd. Funding Information: This prospective study was supported by a grant from Samsung Medison Co. in Seoul, South Korea, which also provided the equipment for this study. The study protocol was reviewed and approved by the Institutional Review Board of Severance hospital. Written informed consent was obtained from all patients before each US examination. All methods were performed in accordance with the relevant guidelines and regulations. Publisher Copyright: {\textcopyright} 2019, The Author(s).",

year = "2019",

month = dec,

day = "1",

doi = "10.1038/s41598-019-54434-1",

language = "English",

volume = "9",

journal = "Scientific reports",

issn = "2045-2322",

publisher = "Nature Publishing Group",

number = "1",

}

TY - JOUR

T1 - Diagnosis of Thyroid Nodules

T2 - Performance of a Deep Learning Convolutional Neural Network Model vs. Radiologists

AU - Park, Vivian Y.

AU - Han, Kyunghwa

AU - Seong, Yeong Kyeong

AU - Park, Moon Ho

AU - Kim, Eun Kyung

AU - Moon, Hee Jung

AU - Yoon, Jung Hyun

AU - Kwak, Jin Young

N1 - Funding Information: This study was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) by the Ministry of Education (2016R1D1A1B03930375) and by the research fund of Samsung Electronics Co., Ltd. Funding Information: This prospective study was supported by a grant from Samsung Medison Co. in Seoul, South Korea, which also provided the equipment for this study. The study protocol was reviewed and approved by the Institutional Review Board of Severance hospital. Written informed consent was obtained from all patients before each US examination. All methods were performed in accordance with the relevant guidelines and regulations. Publisher Copyright: © 2019, The Author(s).

PY - 2019/12/1

Y1 - 2019/12/1

N2 - Computer-aided diagnosis (CAD) systems hold potential to improve the diagnostic accuracy of thyroid ultrasound (US). We aimed to develop a deep learning-based US CAD system (dCAD) for the diagnosis of thyroid nodules and compare its performance with those of a support vector machine (SVM)-based US CAD system (sCAD) and radiologists. dCAD was developed by using US images of 4919 thyroid nodules from three institutions. Its diagnostic performance was prospectively evaluated between June 2016 and February 2017 in 286 nodules, and was compared with those of sCAD and radiologists, using logistic regression with the generalized estimating equation. Subgroup analyses were performed according to experience level and separately for small thyroid nodules 1–2 cm. There was no difference in overall sensitivity, specificity, positive predictive value (PPV), negative predictive value and accuracy (all p > 0.05) between radiologists and dCAD. Radiologists and dCAD showed higher specificity, PPV, and accuracy than sCAD (all p < 0.001). In small nodules, experienced radiologists showed higher specificity, PPV and accuracy than sCAD (all p < 0.05). In conclusion, dCAD showed overall comparable diagnostic performance with radiologists and assessed thyroid nodules more effectively than sCAD, without loss of sensitivity.

AB - Computer-aided diagnosis (CAD) systems hold potential to improve the diagnostic accuracy of thyroid ultrasound (US). We aimed to develop a deep learning-based US CAD system (dCAD) for the diagnosis of thyroid nodules and compare its performance with those of a support vector machine (SVM)-based US CAD system (sCAD) and radiologists. dCAD was developed by using US images of 4919 thyroid nodules from three institutions. Its diagnostic performance was prospectively evaluated between June 2016 and February 2017 in 286 nodules, and was compared with those of sCAD and radiologists, using logistic regression with the generalized estimating equation. Subgroup analyses were performed according to experience level and separately for small thyroid nodules 1–2 cm. There was no difference in overall sensitivity, specificity, positive predictive value (PPV), negative predictive value and accuracy (all p > 0.05) between radiologists and dCAD. Radiologists and dCAD showed higher specificity, PPV, and accuracy than sCAD (all p < 0.001). In small nodules, experienced radiologists showed higher specificity, PPV and accuracy than sCAD (all p < 0.05). In conclusion, dCAD showed overall comparable diagnostic performance with radiologists and assessed thyroid nodules more effectively than sCAD, without loss of sensitivity.

UR - http://www.scopus.com/inward/record.url?scp=85075758502&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85075758502&partnerID=8YFLogxK

U2 - 10.1038/s41598-019-54434-1

DO - 10.1038/s41598-019-54434-1

M3 - Article

C2 - 31780753

AN - SCOPUS:85075758502

SN - 2045-2322

VL - 9

JO - Scientific reports

JF - Scientific reports

IS - 1

M1 - 17843

ER -

Diagnosis of Thyroid Nodules: Performance of a Deep Learning Convolutional Neural Network Model vs. Radiologists

Abstract

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this