Impact of pre-analytical variables on deep learning accuracy in histopathology.

Journal: Histopathology
PMID: 30801768

Abstract

AIMS: Machine learning (ML) binary classification in diagnostic histopathology is an area of intense investigation. Several assumptions, including training image quality/format and the number of training images required, appear to be similar in many studies irrespective of the paucity of supporting evidence. We empirically compared training image file type, training set size, and two common convolutional neural networks (CNNs) using transfer learning (ResNet50 and SqueezeNet).

Authors

  • Andrew D Jones
    Department of Pathology and Laboratory Medicine, University of California Davis Health, Sacramento, CA, USA.
  • John Paul Graff
    Department of Pathology and Laboratory Medicine, University of California Davis Health, Sacramento, CA, USA.
  • Morgan Darrow
    Department of Pathology and Laboratory Medicine, University of California Davis Health, Sacramento, CA, USA.
  • Alexander Borowsky
    Center for Comparative Medicine, University of California, Davis,CA, USA.
  • Kristin A Olson
    Department of Pathology and Laboratory Medicine, University of California Davis Health, Sacramento, CA, USA.
  • Regina Gandour-Edwards
    Department of Pathology and Laboratory Medicine, University of California Davis Health, Sacramento, CA, USA.
  • Ananya Datta Mitra
    Department of Pathology and Laboratory Medicine, University of California Davis Health, Sacramento, CA, USA.
  • Dongguang Wei
    Department of Pathology and Laboratory Medicine, University of California Davis Health, Sacramento, CA, USA.
  • Guofeng Gao
    Department of Pathology and Laboratory Medicine, University of California Davis Health, Sacramento, CA, USA.
  • Blythe Durbin-Johnson
    Division of Biostatistics, University of California Davis, Sacramento, CA, USA.
  • Hooman H Rashidi
    Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania; Computational Pathology and AI Center of Excellence (CPACE), University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania. Electronic address: rashidihh@upmc.edu.

Keywords

External Resources

Popular Topics
Recent Journals