Mid-Level Feature Extractor for Transfer Learning to Small-Scale Dataset of Medical Images

Dong ho Lee; Yeon Lee; Byeong seok Shin

doi:10.1007/978-981-13-9341-9_2

Mid-Level Feature Extractor for Transfer Learning to Small-Scale Dataset of Medical Images

Dong ho Lee, Yeon Lee, Byeong seok Shin

Inha University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

In fine-tuning-based transfer learning, the size of the dataset may affect the learning accuracy. When a dataset scale is small, fine-tuning-based transfer learning methods use high computing costs, similar to a large-scale dataset. we propose a mid-level feature extractor that only retrains the mid-level convolutional layers, resulting in increased efficiency and reduced computing costs. This mid-level feature extractor is likely to provide an effective alternative in training a small-scale medical image dataset. The performance of the mid-level feature extractor is compared with performance of low- and high-level feature extractors, as well as the fine-tuning method. The mid-level feature extractor takes shorter time to converge than other methods, and it shows good accuracy, obtaining an area under the ROC curve (AUC) of 0.87 in untrained test dataset that is very different from training dataset.

Original language	English
Title of host publication	Advances in Computer Science and Ubiquitous Computing, CSA-CUTE 2018
Editors	James J. Park, Doo-Soon Park, Young-Sik Jeong, Yi Pan
Publisher	Springer
Pages	8-13
Number of pages	6
ISBN (Print)	9789811393402
DOIs	https://doi.org/10.1007/978-981-13-9341-9_2
State	Published - 2020
Event	10th International Conference on Computer Science and its Applications, CSA 2018 and the 13th KIPS International Conference on Ubiquitous Information Technologies and Applications, CUTE 2018 - Kuala Lumpre, Malaysia Duration: 17 Dec 2018 → 19 Dec 2018

Publication series

Name	Lecture Notes in Electrical Engineering
Volume	536 LNEE
ISSN (Print)	1876-1100
ISSN (Electronic)	1876-1119

Conference

Conference	10th International Conference on Computer Science and its Applications, CSA 2018 and the 13th KIPS International Conference on Ubiquitous Information Technologies and Applications, CUTE 2018
Country/Territory	Malaysia
City	Kuala Lumpre
Period	17/12/18 → 19/12/18

Bibliographical note

Publisher Copyright:
© 2020, Springer Nature Singapore Pte Ltd.

Keywords

Convolutional neural networks
Machine learning
Medical images
Transfer learning

Access to Document

10.1007/978-981-13-9341-9_2

Cite this

Lee, D. H., Lee, Y., & Shin, B. S. (2020). Mid-Level Feature Extractor for Transfer Learning to Small-Scale Dataset of Medical Images. In J. J. Park, D.-S. Park, Y.-S. Jeong, & Y. Pan (Eds.), Advances in Computer Science and Ubiquitous Computing, CSA-CUTE 2018 (pp. 8-13). (Lecture Notes in Electrical Engineering; Vol. 536 LNEE). Springer. https://doi.org/10.1007/978-981-13-9341-9_2

@inproceedings{f66f9cdbfe9f46169d87d29775b74e6e,

title = "Mid-Level Feature Extractor for Transfer Learning to Small-Scale Dataset of Medical Images",

abstract = "In fine-tuning-based transfer learning, the size of the dataset may affect the learning accuracy. When a dataset scale is small, fine-tuning-based transfer learning methods use high computing costs, similar to a large-scale dataset. we propose a mid-level feature extractor that only retrains the mid-level convolutional layers, resulting in increased efficiency and reduced computing costs. This mid-level feature extractor is likely to provide an effective alternative in training a small-scale medical image dataset. The performance of the mid-level feature extractor is compared with performance of low- and high-level feature extractors, as well as the fine-tuning method. The mid-level feature extractor takes shorter time to converge than other methods, and it shows good accuracy, obtaining an area under the ROC curve (AUC) of 0.87 in untrained test dataset that is very different from training dataset.",

keywords = "Convolutional neural networks, Machine learning, Medical images, Transfer learning",

author = "Lee, {Dong ho} and Yeon Lee and Shin, {Byeong seok}",

note = "Publisher Copyright: {\textcopyright} 2020, Springer Nature Singapore Pte Ltd.; 10th International Conference on Computer Science and its Applications, CSA 2018 and the 13th KIPS International Conference on Ubiquitous Information Technologies and Applications, CUTE 2018 ; Conference date: 17-12-2018 Through 19-12-2018",

year = "2020",

doi = "10.1007/978-981-13-9341-9_2",

language = "English",

isbn = "9789811393402",

series = "Lecture Notes in Electrical Engineering",

publisher = "Springer",

pages = "8--13",

editor = "Park, {James J.} and Doo-Soon Park and Young-Sik Jeong and Yi Pan",

booktitle = "Advances in Computer Science and Ubiquitous Computing, CSA-CUTE 2018",

address = "Germany",

}

Lee, DH, Lee, Y & Shin, BS 2020, Mid-Level Feature Extractor for Transfer Learning to Small-Scale Dataset of Medical Images. in JJ Park, D-S Park, Y-S Jeong & Y Pan (eds), Advances in Computer Science and Ubiquitous Computing, CSA-CUTE 2018. Lecture Notes in Electrical Engineering, vol. 536 LNEE, Springer, pp. 8-13, 10th International Conference on Computer Science and its Applications, CSA 2018 and the 13th KIPS International Conference on Ubiquitous Information Technologies and Applications, CUTE 2018, Kuala Lumpre, Malaysia, 17/12/18. https://doi.org/10.1007/978-981-13-9341-9_2

Mid-Level Feature Extractor for Transfer Learning to Small-Scale Dataset of Medical Images. / Lee, Dong ho; Lee, Yeon; Shin, Byeong seok.
Advances in Computer Science and Ubiquitous Computing, CSA-CUTE 2018. ed. / James J. Park; Doo-Soon Park; Young-Sik Jeong; Yi Pan. Springer, 2020. p. 8-13 (Lecture Notes in Electrical Engineering; Vol. 536 LNEE).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Mid-Level Feature Extractor for Transfer Learning to Small-Scale Dataset of Medical Images

AU - Lee, Dong ho

AU - Lee, Yeon

AU - Shin, Byeong seok

PY - 2020

Y1 - 2020

N2 - In fine-tuning-based transfer learning, the size of the dataset may affect the learning accuracy. When a dataset scale is small, fine-tuning-based transfer learning methods use high computing costs, similar to a large-scale dataset. we propose a mid-level feature extractor that only retrains the mid-level convolutional layers, resulting in increased efficiency and reduced computing costs. This mid-level feature extractor is likely to provide an effective alternative in training a small-scale medical image dataset. The performance of the mid-level feature extractor is compared with performance of low- and high-level feature extractors, as well as the fine-tuning method. The mid-level feature extractor takes shorter time to converge than other methods, and it shows good accuracy, obtaining an area under the ROC curve (AUC) of 0.87 in untrained test dataset that is very different from training dataset.

AB - In fine-tuning-based transfer learning, the size of the dataset may affect the learning accuracy. When a dataset scale is small, fine-tuning-based transfer learning methods use high computing costs, similar to a large-scale dataset. we propose a mid-level feature extractor that only retrains the mid-level convolutional layers, resulting in increased efficiency and reduced computing costs. This mid-level feature extractor is likely to provide an effective alternative in training a small-scale medical image dataset. The performance of the mid-level feature extractor is compared with performance of low- and high-level feature extractors, as well as the fine-tuning method. The mid-level feature extractor takes shorter time to converge than other methods, and it shows good accuracy, obtaining an area under the ROC curve (AUC) of 0.87 in untrained test dataset that is very different from training dataset.

KW - Convolutional neural networks

KW - Machine learning

KW - Medical images

KW - Transfer learning

UR - http://www.scopus.com/inward/record.url?scp=85076854296&partnerID=8YFLogxK

U2 - 10.1007/978-981-13-9341-9_2

DO - 10.1007/978-981-13-9341-9_2

M3 - Conference contribution

AN - SCOPUS:85076854296

SN - 9789811393402

T3 - Lecture Notes in Electrical Engineering

SP - 8

EP - 13

BT - Advances in Computer Science and Ubiquitous Computing, CSA-CUTE 2018

A2 - Park, James J.

A2 - Park, Doo-Soon

A2 - Jeong, Young-Sik

A2 - Pan, Yi

PB - Springer

T2 - 10th International Conference on Computer Science and its Applications, CSA 2018 and the 13th KIPS International Conference on Ubiquitous Information Technologies and Applications, CUTE 2018

Y2 - 17 December 2018 through 19 December 2018

ER -

Mid-Level Feature Extractor for Transfer Learning to Small-Scale Dataset of Medical Images

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this