skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Sparse coding of pathology slides compared to transfer learning with deep neural networks

Journal Article · · BMC Bioinformatics
 [1];  [2];  [1];  [1];  [1]
  1. Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
  2. Rochester Inst. of Technology, Rochester, NY (United States). Chester F. Carlson Center for Imaging Science
+ Show Author Affiliations

Background Histopathology images of tumor biopsies present unique challenges for applying machine learning to the diagnosis and treatment of cancer. The pathology slides are high resolution, often exceeding 1GB, have non-uniform dimensions, and often contain multiple tissue slices of varying sizes surrounded by large empty regions. The locations of abnormal or cancerous cells, which may constitute a small portion of any given tissue sample, are not annotated. Cancer image datasets are also extremely imbalanced, with most slides being associated with relatively common cancers. Since deep representations trained on natural photographs are unlikely to be optimal for classifying pathology slide images, which have different spectral ranges and spatial structure, we here describe an approach for learning features and inferring representations of cancer pathology slides based on sparse coding. Results We show that conventional transfer learning using a state-of-the-art deep learning architecture pre-trained on ImageNet (RESNET) and fine tuned for a binary tumor/no-tumor classification task achieved between 85% and 86% accuracy. However, when all layers up to the last convolutional layer in RESNET are replaced with a single feature map inferred via a sparse coding using a dictionary optimized for sparse reconstruction of unlabeled pathology slides, classification performance improves to over 93%, corresponding to a 54% error reduction. Conclusions We conclude that a feature dictionary optimized for biomedical imagery may in general support better classification performance than does conventional transfer learning using a dictionary pre-trained on natural images.

Research Organization:
Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
Sponsoring Organization:
USDOE Office of Science (SC)
Grant/Contract Number:
AC52-06NA25396
OSTI ID:
1626772
Journal Information:
BMC Bioinformatics, Vol. 19, Issue S18; ISSN 1471-2105
Publisher:
BioMed CentralCopyright Statement
Country of Publication:
United States
Language:
English

References (31)

Stacked Predictive Sparse Coding for Classification of Distinct Regions in Tumor Histopathology conference December 2013
A Deconvolutional Strategy for Implementing Large Patch Sizes Supports Improved Image Classification conference January 2016
Deep Learning in Medical Image Analysis journal June 2017
Deep learning journal May 2015
Sparse Coding via Thresholding and Local Competition in Neural Circuits journal October 2008
Cancer Genome Landscapes journal March 2013
Large scale tissue histopathology image classification, segmentation, and visualization via deep convolutional activation features journal May 2017
A survey on deep learning in medical image analysis journal December 2017
Deep convolutional neural networks for automatic classification of gastric carcinoma using whole slide images in digital histopathology journal November 2017
Emergence of simple-cell receptive field properties by learning a sparse code for natural images journal June 1996
OpenSlide: A vendor-neutral software foundation for digital pathology journal January 2013
Robust uncertainty principles: exact signal reconstruction from highly incomplete frequency information journal February 2006
Convolutional deep belief network with feature encoding for classification of neuroblastoma histological images journal January 2018
Deep Convolutional Neural Networks Enable Discrimination of Heterogeneous Digital Pathology Images journal January 2018
Convolutional Neural Networks for Medical Image Analysis: Full Training or Fine Tuning? journal May 2016
Digital image analysis in breast pathology—from image processing techniques to artificial intelligence journal April 2018
Compressed sensing journal April 2006
A Threshold Selection Method from Gray-Level Histograms journal January 1979
SPORCO: A Python package for standard and convolutional sparse representations conference January 2017
ImageNet Large Scale Visual Recognition Challenge journal April 2015
OpenSlide: A vendor-neutral software foundation for digital pathology. text January 2002
ImageNet Large Scale Visual Recognition Challenge text January 2015
OpenSlide: A vendor-neutral software foundation for digital pathology. text January 2002
Deep Learning in Medical Image Analysis journal April 2021
Deep Learning text January 2018
Deep Learning in Medical Image Analysis text January 2017
A Survey on Deep Learning in Medical Image Analysis text January 2017
Robust Uncertainty Principles: Exact Signal Reconstruction from Highly Incomplete Frequency Information preprint January 2004
Polycystic liver: automatic segmentation using deep learning on CT is faster and as accurate compared to manual segmentation journal February 2022
On some common compressive sensing recovery algorithms and applications - Review paper preprint January 2017
Prognostic Value of Transfer Learning Based Features in Resectable Pancreatic Ductal Adenocarcinoma preprint January 2019

Cited By (2)

Extracting Crop Spatial Distribution from Gaofen 2 Imagery Using a Convolutional Neural Network journal July 2019
A New CNN-Bayesian Model for Extracting Improved Winter Wheat Spatial Distribution from GF-2 imagery journal March 2019

Linked Research (4)

Figures / Tables (7)

Table 1(p. 3)table Table 1
Fig. 1(p. 4)figure Fig. 1
Fig. 2(p. 5)figure Fig. 2
Fig. 3(p. 5)figure Fig. 3
Fig. 4(p. 6)figure Fig. 4
Fig. 5(p. 7)figure Fig. 5
Table 2(p. 7)table Table 2

Similar Records

Related Subjects

59 BASIC BIOLOGICAL SCIENCES
Biochemistry & Molecular Biology
Biotechnology & Applied Microbiology
Mathematical & Computational Biology