Unsupervised Representation Learning from Pathology Images with Multi-directional Contrastive Predictive Coding

TY - GEN

T1 - Unsupervised Representation Learning from Pathology Images with Multi-directional Contrastive Predictive Coding

AU - Carse, Jacob

AU - Carey, Frank

AU - McKenna, Stephen

N1 - Copyright: © 2021 IEEE.

PY - 2021/5/25

Y1 - 2021/5/25

N2 - Digital pathology tasks have benefited greatly from modern deep learning algorithms. However, their need for large quantities of annotated data has been identified as a key challenge. This need for data can be countered by using unsupervised learning in situations where data are abundant but access to annotations is limited. Feature representations learned from unannotated data using contrastive predictive coding (CPC) have been shown to enable classifiers to obtain state of the art performance from relatively small amounts of annotated computer vision data. We present a modification to the CPC framework for use with digital pathology patches. This is achieved by introducing an alternative mask for building the latent context and using a multi-directional PixelCNN autoregressor. To demonstrate our proposed method we learn feature representations from the Patch Camelyon histology dataset. We show that our proposed modification can yield improved deep classification of histology patches.

AB - Digital pathology tasks have benefited greatly from modern deep learning algorithms. However, their need for large quantities of annotated data has been identified as a key challenge. This need for data can be countered by using unsupervised learning in situations where data are abundant but access to annotations is limited. Feature representations learned from unannotated data using contrastive predictive coding (CPC) have been shown to enable classifiers to obtain state of the art performance from relatively small amounts of annotated computer vision data. We present a modification to the CPC framework for use with digital pathology patches. This is achieved by introducing an alternative mask for building the latent context and using a multi-directional PixelCNN autoregressor. To demonstrate our proposed method we learn feature representations from the Patch Camelyon histology dataset. We show that our proposed modification can yield improved deep classification of histology patches.

KW - Digital Pathology

KW - Medical Imaging

KW - Representation Learning

KW - Semi-Supervised Learning

UR - https://arxiv.org/abs/2105.05345

UR - https://virtual.biomedicalimaging.org/isbi/2021/isbi-2021/315261/jacob.carse.unsupervised.representation.learning.from.pathology.images.with.html?f=listing%3D1%2Abrowseby%3D8%2Asortby%3D2%2Amedia%3D2%2Ace_id%3D1979%2Aot_id%3D24333

UR - https://www.scopus.com/pages/publications/85107183330

U2 - 10.1109/ISBI48211.2021.9434140

DO - 10.1109/ISBI48211.2021.9434140

M3 - Conference contribution

SN - 9781665429474

T3 - Proceedings - International Symposium on Biomedical Imaging

SP - 1254

EP - 1258

BT - 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI 2021)

PB - IEEE

T2 - 2021 IEEE International Symposium on Biomedical Imaging (ISBI)

Y2 - 13 April 2021 through 16 April 2021

ER -