Advanced search
2 files | 1.11 MB Add to list

Self-supervised learning as a means to reduce the need for labeled data in medical image analysis

Author
Organization
Abstract
One of the largest problems in medical image processing is the lack of annotated data. Labeling medical images often requires highly trained experts and can be a time-consuming process. In this paper, we evaluate a method of reducing the need for labeled data in medical image object detection by using self-supervised neural network pretraining. We use a dataset of chest X-ray images with bounding box labels for 13 different classes of anomalies. The networks are pretrained on a percentage of the dataset without labels and then fine-tuned on the rest of the dataset. We show that it is possible to achieve similar performance to a fully supervised model in terms of mean average precision and accuracy with only 60% of the labeled data. We also show that it is possible to increase the maximum performance of a fully-supervised model by adding a self-supervised pretraining step, and this effect can be observed with even a small amount of unlabeled data for pretraining.
Keywords
Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, contrastive learning, deep learning, medical image processing, object detection, self-supervised learning

Downloads

  • 2206.00344.pdf
    • full text (Author's original)
    • |
    • open access
    • |
    • PDF
    • |
    • 555.89 KB
  • (...).pdf
    • full text (Published version)
    • |
    • UGent only
    • |
    • PDF
    • |
    • 553.57 KB

Citation

Please use this url to cite or link to this publication:

MLA
Benčević, Marin, et al. “Self-Supervised Learning as a Means to Reduce the Need for Labeled Data in Medical Image Analysis.” 2022 30th European Signal Processing Conference (EUSIPCO), IEEE, 2022, pp. 1328–32, doi:10.23919/EUSIPCO55093.2022.9909542.
APA
Benčević, M., Habijan, M., Galić, I., & Pizurica, A. (2022). Self-supervised learning as a means to reduce the need for labeled data in medical image analysis. 2022 30th European Signal Processing Conference (EUSIPCO), 1328–1332. https://doi.org/10.23919/EUSIPCO55093.2022.9909542
Chicago author-date
Benčević, Marin, Marija Habijan, Irena Galić, and Aleksandra Pizurica. 2022. “Self-Supervised Learning as a Means to Reduce the Need for Labeled Data in Medical Image Analysis.” In 2022 30th European Signal Processing Conference (EUSIPCO), 1328–32. IEEE. https://doi.org/10.23919/EUSIPCO55093.2022.9909542.
Chicago author-date (all authors)
Benčević, Marin, Marija Habijan, Irena Galić, and Aleksandra Pizurica. 2022. “Self-Supervised Learning as a Means to Reduce the Need for Labeled Data in Medical Image Analysis.” In 2022 30th European Signal Processing Conference (EUSIPCO), 1328–1332. IEEE. doi:10.23919/EUSIPCO55093.2022.9909542.
Vancouver
1.
Benčević M, Habijan M, Galić I, Pizurica A. Self-supervised learning as a means to reduce the need for labeled data in medical image analysis. In: 2022 30th European Signal Processing Conference (EUSIPCO). IEEE; 2022. p. 1328–32.
IEEE
[1]
M. Benčević, M. Habijan, I. Galić, and A. Pizurica, “Self-supervised learning as a means to reduce the need for labeled data in medical image analysis,” in 2022 30th European Signal Processing Conference (EUSIPCO), Belgrade, Serbia, 2022, pp. 1328–1332.
@inproceedings{8760360,
  abstract     = {{One of the largest problems in medical image processing is the lack of annotated data. Labeling medical images often requires highly trained experts and can be a time-consuming process. In this paper, we evaluate a method of reducing the need for labeled data in medical image object detection by using self-supervised neural network pretraining. We use a dataset of chest X-ray images with bounding box labels for 13 different classes of anomalies. The networks are pretrained on a percentage of the dataset without labels and then fine-tuned on the rest of the dataset. We show that it is possible to achieve similar performance to a fully supervised model in terms of mean average precision and accuracy with only 60% of the labeled data. We also show that it is possible to increase the maximum performance of a fully-supervised model by adding a self-supervised pretraining step, and this effect can be observed with even a small amount of unlabeled data for pretraining.}},
  author       = {{Benčević, Marin and Habijan, Marija and Galić, Irena and Pizurica, Aleksandra}},
  booktitle    = {{2022 30th European Signal Processing Conference (EUSIPCO)}},
  isbn         = {{9789082797091}},
  issn         = {{2076-1465}},
  keywords     = {{Computer Vision and Pattern Recognition (cs.CV),FOS: Computer and information sciences,contrastive learning,deep learning,medical image processing,object detection,self-supervised learning}},
  language     = {{eng}},
  location     = {{Belgrade, Serbia}},
  pages        = {{1328--1332}},
  publisher    = {{IEEE}},
  title        = {{Self-supervised learning as a means to reduce the need for labeled data in medical image analysis}},
  url          = {{http://doi.org/10.23919/EUSIPCO55093.2022.9909542}},
  year         = {{2022}},
}

Altmetric
View in Altmetric
Web of Science
Times cited: