UKSSL: Underlying Knowledge based Semi-Supervised Learning for Medical Image Classification

Zeyu Ren, Xiangyu Kong, Yudong Zhang, Shuihua Wang

Research output: Contribution to journalArticlepeer-review

39 Citations (Scopus)

Abstract

<italic>Goal:</italic> Deep learning techniques have made significant progress in medical image analysis. However, obtaining ground truth labels for unlabeled medical images is challenging as they often outnumber labeled images. Thus, training a high-performance model with limited labeled data has become a crucial challenge. <italic>Methods:</italic> This study introduces an underlying knowledge-based semi-supervised framework called UKSSL, consisting of two components: MedCLR extracts feature representations from the unlabeled dataset; UKMLP utilizes the representation and fine-tunes it with the limited labeled dataset to classify the medical images. <italic>Results:</italic> UKSSL evaluates on the LC25000 and BCCD datasets, using only 50&#x0025; labeled data. It gets precision, recall, F1-score, and accuracy of 98.9&#x0025; on LC25000 and 94.3&#x0025;, 94.5&#x0025;, 94.3&#x0025;, and 94.1&#x0025; on BCCD, respectively. These results outperform other supervised-learning methods using 100&#x0025; labeled data. <italic>Conclusions</italic>: The UKSSL can efficiently extract underlying knowledge from the unlabeled dataset and perform better using limited labeled medical images.

Original languageEnglish
Pages (from-to)1-8
Number of pages8
JournalIEEE Open Journal of Engineering in Medicine and Biology
DOIs
Publication statusAccepted/In press - 2023
Externally publishedYes

Keywords

  • Biomedical imaging
  • Deep Learning
  • Head
  • Image Classification
  • Image augmentation
  • Medical Image Analysis
  • Self-supervised Learning
  • Self-supervised learning
  • Semantics
  • Semi-supervised Learning
  • Semisupervised learning
  • Task analysis

Fingerprint

Dive into the research topics of 'UKSSL: Underlying Knowledge based Semi-Supervised Learning for Medical Image Classification'. Together they form a unique fingerprint.

Cite this