September 2019
Volume 19, Issue 10
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2019
Perceptual learning of chest X-ray images
Author Affiliations & Notes
  • Sha Li
    Department of Psychology, University of Minnesota
  • Roger W Remington
    School of Psychology, University of Queensland
    Center for Cognitive Sciences, University of Minnesota
  • Yuhong V Jiang
    Department of Psychology, University of Minnesota
    Center for Cognitive Sciences, University of Minnesota
Journal of Vision September 2019, Vol.19, 28a. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Sha Li, Roger W Remington, Yuhong V Jiang; Perceptual learning of chest X-ray images. Journal of Vision 2019;19(10):28a. doi:

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Extensive research on perceptual learning has shown that adults can readily acquire new perceptual skills. Much of this research, however, has focused on simple properties such as contrast or orientation, and training yields limited transfer to new features or locations. Here we investigated the acquisition and transfer of more complex perceptual skills. We trained novices to identify lung cancer and examined two types of learning: local learning of tumor properties and global learning of the surrounding context. Stimuli were taken from the Japanese Society of Radiological Technology and included both abnormal and normal chest X-rays. Participants underwent 4 days of training (180 trials each), during which they saw a set of 60 images presented multiple times. On each training trial, participants saw a pair of chest X-ray images, one abnormal and one normal, and were asked to choose the abnormal image and localize the tumor. The computer provided feedback regarding the correct tumor location. Both classification and localization accuracy improved across training sessions, indicating the occurrence of learning. We also administered a pre-test before the first training session and a post-test after each training session. The test contained a single image for participants to classify as cancerous or not. It included both trained images and new images, presented in one of three formats – the entire image, just the cutout of the cancerous region or a comparable region in normal images, or just the surrounding region. For trained images, participants were able to classify all three formats at above-chance levels. Learning transferred to untrained images, especially when the entire image was available for classification. Accuracy for classifying just the cutout or the surrounding regions of untrained images was low but above chance. These results showed that moderate amount of training yields transferrable perceptual learning of complex visual stimuli.

Acknowledgement: This work was supported by a seed grant from OFAA-Social Sciences, University of Minnesota to YVJ and Doctoral Dissertation Research Award from American Psychological Association and Doctoral Dissertation Fellowship from the University of Minnesota to SL. 

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.