Affordable Access

Publisher Website

A Novel Distant Domain Transfer Learning Framework for Thyroid Image Classification.

Authors
  • Tang, Fenghe1
  • Ding, Jianrui1
  • Wang, Lingtao1
  • Ning, Chunping2
  • 1 School of Computer Science and Technology, Harbin Institute of Technology, Harbin, China. , (China)
  • 2 Ultrasound Department, The Affiliated Hospital of Qingdao University, Qingdao, China. , (China)
Type
Published Article
Journal
Neural Processing Letters
Publisher
Springer Science and Business Media LLC
Publication Date
Jun 25, 2022
Pages
1–17
Identifiers
DOI: 10.1007/s11063-022-10940-4
PMID: 35789884
Source
Medline
Keywords
Language
English
License
Unknown

Abstract

Medical ultrasound imaging technology is currently the preferred method for early diagnosis of thyroid nodules. Radiologists' analysis of ultrasound images is highly dependent on their clinical experience and is susceptible to intra- and inter-observer variability. Although end-to-end deep learning technique can address these limitations, the difficulty of acquiring annotated medical image makes it very challenging. Transfer learning can alleviate the problems, but the large gap between source and target domain will lead to negative transfer. In this paper, a novel transfer learning method with distant domain high-level feature fusion (DHFF) model is proposed. It reduces the distribution distance between the source domain and the target domain while maintaining the characteristics of respective domains, which can avoid excessive feature fusion while enabling the model to learn more valuable transfer knowledge. The DHFF is validated by multiple public source and private target datasets in experiments. The results show that the classification accuracy of DHFF is up to 88.92% with thyroid ultrasound auxiliary source domains, which is up to 8% higher than existing transfer and distant transfer algorithms. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022.

Report this publication

Statistics

Seen <100 times