Classification of tumor histopathology via sparse feature learning

Nandita Nayak, Hang Chang, Alexander Borowsky, Paul Spellman, Bahram Parvin

Research output: Chapter in Book/Report/Conference proceedingConference contribution

12 Citations (Scopus)

Abstract

Our goal is to decompose whole slide images (WSI) of histology sections into distinct patches (e.g., viable tumor, necrosis) so that statistics of distinct histopathology can be linked with the outcome. Such an analysis requires a large cohort of histology sections that may originate from different laboratories, which may not use the same protocol in sample preparation. We have evaluated a method based on a variation of the restricted Boltzmann machine (RBM) that learns intrinsic features of the image signature in an unsupervised fashion. Computed code, from the learned representation, is then utilized to classify patches from a curated library of images. The system has been evaluated against a dataset of small image blocks of 1k-by-1k that have been extracted from glioblastoma multiforme (GBM) and clear cell kidney carcinoma (KIRC) from the cancer genome atlas (TCGA) archive. The learned model is then projected on each whole slide image (e.g., of size 20k-by-20k pixels or larger) for characterizing and visualizing tumor architecture. In the case of GBM, each WSI is decomposed into necrotic, transition into necrosis, and viable. In the case of the KIRC, each WSI is decomposed into tumor types, stroma, normal, and others. Evaluation of 1400 and 2500 samples of GBM and KIRC indicates a performance of 84% and 81%, respectively.

Original languageEnglish (US)
Title of host publicationProceedings - International Symposium on Biomedical Imaging
Pages410-413
Number of pages4
DOIs
StatePublished - 2013
Externally publishedYes
Event2013 IEEE 10th International Symposium on Biomedical Imaging: From Nano to Macro, ISBI 2013 - San Francisco, CA, United States
Duration: Apr 7 2013Apr 11 2013

Other

Other2013 IEEE 10th International Symposium on Biomedical Imaging: From Nano to Macro, ISBI 2013
CountryUnited States
CitySan Francisco, CA
Period4/7/134/11/13

Fingerprint

Glioblastoma
Tumors
Histology
Learning
Carcinoma
Necrosis
Kidney
Neoplasms
Atlases
Kidney Neoplasms
Libraries
Genes
Pixels
Statistics
Genome

Keywords

  • feature learning
  • sparse coding
  • Tumor characterization
  • whole slide imaging

ASJC Scopus subject areas

  • Biomedical Engineering
  • Radiology Nuclear Medicine and imaging

Cite this

Nayak, N., Chang, H., Borowsky, A., Spellman, P., & Parvin, B. (2013). Classification of tumor histopathology via sparse feature learning. In Proceedings - International Symposium on Biomedical Imaging (pp. 410-413). [6556499] https://doi.org/10.1109/ISBI.2013.6556499

Classification of tumor histopathology via sparse feature learning. / Nayak, Nandita; Chang, Hang; Borowsky, Alexander; Spellman, Paul; Parvin, Bahram.

Proceedings - International Symposium on Biomedical Imaging. 2013. p. 410-413 6556499.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Nayak, N, Chang, H, Borowsky, A, Spellman, P & Parvin, B 2013, Classification of tumor histopathology via sparse feature learning. in Proceedings - International Symposium on Biomedical Imaging., 6556499, pp. 410-413, 2013 IEEE 10th International Symposium on Biomedical Imaging: From Nano to Macro, ISBI 2013, San Francisco, CA, United States, 4/7/13. https://doi.org/10.1109/ISBI.2013.6556499
Nayak N, Chang H, Borowsky A, Spellman P, Parvin B. Classification of tumor histopathology via sparse feature learning. In Proceedings - International Symposium on Biomedical Imaging. 2013. p. 410-413. 6556499 https://doi.org/10.1109/ISBI.2013.6556499
Nayak, Nandita ; Chang, Hang ; Borowsky, Alexander ; Spellman, Paul ; Parvin, Bahram. / Classification of tumor histopathology via sparse feature learning. Proceedings - International Symposium on Biomedical Imaging. 2013. pp. 410-413
@inproceedings{f14a225cf9f44c7a8976b4eaebfa30d9,
title = "Classification of tumor histopathology via sparse feature learning",
abstract = "Our goal is to decompose whole slide images (WSI) of histology sections into distinct patches (e.g., viable tumor, necrosis) so that statistics of distinct histopathology can be linked with the outcome. Such an analysis requires a large cohort of histology sections that may originate from different laboratories, which may not use the same protocol in sample preparation. We have evaluated a method based on a variation of the restricted Boltzmann machine (RBM) that learns intrinsic features of the image signature in an unsupervised fashion. Computed code, from the learned representation, is then utilized to classify patches from a curated library of images. The system has been evaluated against a dataset of small image blocks of 1k-by-1k that have been extracted from glioblastoma multiforme (GBM) and clear cell kidney carcinoma (KIRC) from the cancer genome atlas (TCGA) archive. The learned model is then projected on each whole slide image (e.g., of size 20k-by-20k pixels or larger) for characterizing and visualizing tumor architecture. In the case of GBM, each WSI is decomposed into necrotic, transition into necrosis, and viable. In the case of the KIRC, each WSI is decomposed into tumor types, stroma, normal, and others. Evaluation of 1400 and 2500 samples of GBM and KIRC indicates a performance of 84{\%} and 81{\%}, respectively.",
keywords = "feature learning, sparse coding, Tumor characterization, whole slide imaging",
author = "Nandita Nayak and Hang Chang and Alexander Borowsky and Paul Spellman and Bahram Parvin",
year = "2013",
doi = "10.1109/ISBI.2013.6556499",
language = "English (US)",
isbn = "9781467364546",
pages = "410--413",
booktitle = "Proceedings - International Symposium on Biomedical Imaging",

}

TY - GEN

T1 - Classification of tumor histopathology via sparse feature learning

AU - Nayak, Nandita

AU - Chang, Hang

AU - Borowsky, Alexander

AU - Spellman, Paul

AU - Parvin, Bahram

PY - 2013

Y1 - 2013

N2 - Our goal is to decompose whole slide images (WSI) of histology sections into distinct patches (e.g., viable tumor, necrosis) so that statistics of distinct histopathology can be linked with the outcome. Such an analysis requires a large cohort of histology sections that may originate from different laboratories, which may not use the same protocol in sample preparation. We have evaluated a method based on a variation of the restricted Boltzmann machine (RBM) that learns intrinsic features of the image signature in an unsupervised fashion. Computed code, from the learned representation, is then utilized to classify patches from a curated library of images. The system has been evaluated against a dataset of small image blocks of 1k-by-1k that have been extracted from glioblastoma multiforme (GBM) and clear cell kidney carcinoma (KIRC) from the cancer genome atlas (TCGA) archive. The learned model is then projected on each whole slide image (e.g., of size 20k-by-20k pixels or larger) for characterizing and visualizing tumor architecture. In the case of GBM, each WSI is decomposed into necrotic, transition into necrosis, and viable. In the case of the KIRC, each WSI is decomposed into tumor types, stroma, normal, and others. Evaluation of 1400 and 2500 samples of GBM and KIRC indicates a performance of 84% and 81%, respectively.

AB - Our goal is to decompose whole slide images (WSI) of histology sections into distinct patches (e.g., viable tumor, necrosis) so that statistics of distinct histopathology can be linked with the outcome. Such an analysis requires a large cohort of histology sections that may originate from different laboratories, which may not use the same protocol in sample preparation. We have evaluated a method based on a variation of the restricted Boltzmann machine (RBM) that learns intrinsic features of the image signature in an unsupervised fashion. Computed code, from the learned representation, is then utilized to classify patches from a curated library of images. The system has been evaluated against a dataset of small image blocks of 1k-by-1k that have been extracted from glioblastoma multiforme (GBM) and clear cell kidney carcinoma (KIRC) from the cancer genome atlas (TCGA) archive. The learned model is then projected on each whole slide image (e.g., of size 20k-by-20k pixels or larger) for characterizing and visualizing tumor architecture. In the case of GBM, each WSI is decomposed into necrotic, transition into necrosis, and viable. In the case of the KIRC, each WSI is decomposed into tumor types, stroma, normal, and others. Evaluation of 1400 and 2500 samples of GBM and KIRC indicates a performance of 84% and 81%, respectively.

KW - feature learning

KW - sparse coding

KW - Tumor characterization

KW - whole slide imaging

UR - http://www.scopus.com/inward/record.url?scp=84881642017&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84881642017&partnerID=8YFLogxK

U2 - 10.1109/ISBI.2013.6556499

DO - 10.1109/ISBI.2013.6556499

M3 - Conference contribution

AN - SCOPUS:84881642017

SN - 9781467364546

SP - 410

EP - 413

BT - Proceedings - International Symposium on Biomedical Imaging

ER -