When machine vision meets histology: A comparative evaluation of model architecture for classification of histology sections

Cheng Zhong, Ju Han, Alexander D Borowsky, Bahram Parvin, Yunfu Wang, Hang Chang

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

Classification of histology sections in large cohorts, in terms of distinct regions of microanatomy (e.g., stromal) and histopathology (e.g., tumor, necrosis), enables the quantification of tumor composition, and the construction of predictive models of genomics and clinical outcome. To tackle the large technical variations and biological heterogeneities, which are intrinsic in large cohorts, emerging systems utilize either prior knowledge from pathologists or unsupervised feature learning for invariant representation of the underlying properties in the data. However, to a large degree, the architecture for tissue histology classification remains unexplored and requires urgent systematical investigation. This paper is the first attempt to provide insights into three fundamental questions in tissue histology classification: I. Is unsupervised feature learning preferable to human engineered features? II. Does cellular saliency help? III. Does the sparse feature encoder contribute to recognition? We show that (a) in I, both Cellular Morphometric Feature and features from unsupervised feature learning lead to superior performance when compared to SIFT and [Color, Texture]; (b) in II, cellular saliency incorporation impairs the performance for systems built upon pixel-/patch-level features; and (c) in III, the effect of the sparse feature encoder is correlated with the robustness of features, and the performance can be consistently improved by the multi-stage extension of systems built upon both Cellular Morphmetric Feature and features from unsupervised feature learning. These insights are validated with two cohorts of Glioblastoma Multiforme (GBM) and Kidney Clear Cell Carcinoma (KIRC).

Original languageEnglish (US)
Pages (from-to)530-543
Number of pages14
JournalMedical Image Analysis
Volume35
DOIs
StatePublished - Jan 1 2017

Fingerprint

Comparative Histology
Histology
Computer vision
Learning
Tumors
Tissue
Glioblastoma
Textures
Genomics
Pixels
Cells
Color
Neoplasms
Necrosis
Carcinoma
Chemical analysis
Kidney

Keywords

  • Classification
  • Computational histopathology
  • Sparse feature encoder
  • Unsupervised feature learning

ASJC Scopus subject areas

  • Radiological and Ultrasound Technology
  • Radiology Nuclear Medicine and imaging
  • Computer Vision and Pattern Recognition
  • Health Informatics
  • Computer Graphics and Computer-Aided Design

Cite this

When machine vision meets histology : A comparative evaluation of model architecture for classification of histology sections. / Zhong, Cheng; Han, Ju; Borowsky, Alexander D; Parvin, Bahram; Wang, Yunfu; Chang, Hang.

In: Medical Image Analysis, Vol. 35, 01.01.2017, p. 530-543.

Research output: Contribution to journalArticle

@article{5595c9bba0634b1cb3cd3565884645ed,
title = "When machine vision meets histology: A comparative evaluation of model architecture for classification of histology sections",
abstract = "Classification of histology sections in large cohorts, in terms of distinct regions of microanatomy (e.g., stromal) and histopathology (e.g., tumor, necrosis), enables the quantification of tumor composition, and the construction of predictive models of genomics and clinical outcome. To tackle the large technical variations and biological heterogeneities, which are intrinsic in large cohorts, emerging systems utilize either prior knowledge from pathologists or unsupervised feature learning for invariant representation of the underlying properties in the data. However, to a large degree, the architecture for tissue histology classification remains unexplored and requires urgent systematical investigation. This paper is the first attempt to provide insights into three fundamental questions in tissue histology classification: I. Is unsupervised feature learning preferable to human engineered features? II. Does cellular saliency help? III. Does the sparse feature encoder contribute to recognition? We show that (a) in I, both Cellular Morphometric Feature and features from unsupervised feature learning lead to superior performance when compared to SIFT and [Color, Texture]; (b) in II, cellular saliency incorporation impairs the performance for systems built upon pixel-/patch-level features; and (c) in III, the effect of the sparse feature encoder is correlated with the robustness of features, and the performance can be consistently improved by the multi-stage extension of systems built upon both Cellular Morphmetric Feature and features from unsupervised feature learning. These insights are validated with two cohorts of Glioblastoma Multiforme (GBM) and Kidney Clear Cell Carcinoma (KIRC).",
keywords = "Classification, Computational histopathology, Sparse feature encoder, Unsupervised feature learning",
author = "Cheng Zhong and Ju Han and Borowsky, {Alexander D} and Bahram Parvin and Yunfu Wang and Hang Chang",
year = "2017",
month = "1",
day = "1",
doi = "10.1016/j.media.2016.08.010",
language = "English (US)",
volume = "35",
pages = "530--543",
journal = "Medical Image Analysis",
issn = "1361-8415",
publisher = "Elsevier",

}

TY - JOUR

T1 - When machine vision meets histology

T2 - A comparative evaluation of model architecture for classification of histology sections

AU - Zhong, Cheng

AU - Han, Ju

AU - Borowsky, Alexander D

AU - Parvin, Bahram

AU - Wang, Yunfu

AU - Chang, Hang

PY - 2017/1/1

Y1 - 2017/1/1

N2 - Classification of histology sections in large cohorts, in terms of distinct regions of microanatomy (e.g., stromal) and histopathology (e.g., tumor, necrosis), enables the quantification of tumor composition, and the construction of predictive models of genomics and clinical outcome. To tackle the large technical variations and biological heterogeneities, which are intrinsic in large cohorts, emerging systems utilize either prior knowledge from pathologists or unsupervised feature learning for invariant representation of the underlying properties in the data. However, to a large degree, the architecture for tissue histology classification remains unexplored and requires urgent systematical investigation. This paper is the first attempt to provide insights into three fundamental questions in tissue histology classification: I. Is unsupervised feature learning preferable to human engineered features? II. Does cellular saliency help? III. Does the sparse feature encoder contribute to recognition? We show that (a) in I, both Cellular Morphometric Feature and features from unsupervised feature learning lead to superior performance when compared to SIFT and [Color, Texture]; (b) in II, cellular saliency incorporation impairs the performance for systems built upon pixel-/patch-level features; and (c) in III, the effect of the sparse feature encoder is correlated with the robustness of features, and the performance can be consistently improved by the multi-stage extension of systems built upon both Cellular Morphmetric Feature and features from unsupervised feature learning. These insights are validated with two cohorts of Glioblastoma Multiforme (GBM) and Kidney Clear Cell Carcinoma (KIRC).

AB - Classification of histology sections in large cohorts, in terms of distinct regions of microanatomy (e.g., stromal) and histopathology (e.g., tumor, necrosis), enables the quantification of tumor composition, and the construction of predictive models of genomics and clinical outcome. To tackle the large technical variations and biological heterogeneities, which are intrinsic in large cohorts, emerging systems utilize either prior knowledge from pathologists or unsupervised feature learning for invariant representation of the underlying properties in the data. However, to a large degree, the architecture for tissue histology classification remains unexplored and requires urgent systematical investigation. This paper is the first attempt to provide insights into three fundamental questions in tissue histology classification: I. Is unsupervised feature learning preferable to human engineered features? II. Does cellular saliency help? III. Does the sparse feature encoder contribute to recognition? We show that (a) in I, both Cellular Morphometric Feature and features from unsupervised feature learning lead to superior performance when compared to SIFT and [Color, Texture]; (b) in II, cellular saliency incorporation impairs the performance for systems built upon pixel-/patch-level features; and (c) in III, the effect of the sparse feature encoder is correlated with the robustness of features, and the performance can be consistently improved by the multi-stage extension of systems built upon both Cellular Morphmetric Feature and features from unsupervised feature learning. These insights are validated with two cohorts of Glioblastoma Multiforme (GBM) and Kidney Clear Cell Carcinoma (KIRC).

KW - Classification

KW - Computational histopathology

KW - Sparse feature encoder

KW - Unsupervised feature learning

UR - http://www.scopus.com/inward/record.url?scp=84987984338&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84987984338&partnerID=8YFLogxK

U2 - 10.1016/j.media.2016.08.010

DO - 10.1016/j.media.2016.08.010

M3 - Article

C2 - 27644083

AN - SCOPUS:84987984338

VL - 35

SP - 530

EP - 543

JO - Medical Image Analysis

JF - Medical Image Analysis

SN - 1361-8415

ER -