Using Transfer Learning with Convolutional Neural Networks to Diagnose Breast Cancer from Histopathological Images

Weiming Zhi, Henry Wing Fung Yueng, Zhenghao Chen, Seid Miad Zandavi, Zhicheng Lu, Yuk Ying Chung

Research output: Book chapter/Published conference paperConference paperpeer-review

46 Citations (Scopus)

Abstract

Diagnosis from histopathological images is the gold standard in diagnosing breast cancer. This paper investigates using transfer learning with convolutional neural networks to automatically diagnose breast cancer from patches of histopathological images. We compare the performance of using transfer learning with an off-the-shelf deep convolutional neural network architecture, VGGNet, and a shallower custom architecture. Our proposed final ensemble model, which contains three custom convolutional neural network classifiers trained using transfer learning, achieves a significantly higher image classification accuracy on the large public benchmark dataset than the current best results, for all image resolution levels.

Original languageEnglish
Title of host publicationNeural Information Processing - 24th International Conference, ICONIP 2017, Proceedings
EditorsDerong Liu, Shengli Xie, Yuanqing Li, El-Sayed M. El-Alfy, Dongbin Zhao
PublisherSpringer-Verlag Italia Srl
Pages669-676
Number of pages8
ISBN (Print)9783319700922
DOIs
Publication statusPublished - 2017
Event24th International Conference on Neural Information Processing, ICONIP 2017 - Guangzhou, China
Duration: 14 Nov 201718 Nov 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10637 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference24th International Conference on Neural Information Processing, ICONIP 2017
Country/TerritoryChina
CityGuangzhou
Period14/11/1718/11/17

Fingerprint

Dive into the research topics of 'Using Transfer Learning with Convolutional Neural Networks to Diagnose Breast Cancer from Histopathological Images'. Together they form a unique fingerprint.

Cite this