DeepFeature: Unsupervised Feature Extraction via Deep Learning for Histopathological Images

This code is the implementation of an unsupervised feature extractor called DeepFeature that we proposed for effective representation and classification of histopathological tissue images. This feature extractor characterizes a tissue image by first identifying its salient subregions and then using only the quantizations of these subregions. For the characterization of a subregion, it feeds the subregion’s pixels to a deep belief network of consecutive RBMs and defines the activation values of the hidden units in the last RBM layer as the deep features of this subregion. Then, it clusters these deep features to learn the quantizations in an unsupervised way.

The source codes are provided here.

NOTE: The following source codes are provided for research purposes only. The authors have no responsibility for any consequences of use of these source codes. If you use any part of these codes, please cite the following paper.

  • C. T. Sari and C. Gunduz-Demir, “Unsupervised feature extraction via deep learning for histopathological classification of colon tissue images, IEEE Transactions on Medical Imaging, 38(5):1139-1149, 2019.

Please contact Can Taylan Sari for further questions.

Source code

The program is available for Matlab. Read the howToRun.m file for the use of this code. This file contains functions that use Matlab codes as well as C codes, which should first be compiled by the mex compiler in Matlab. The howToRun.m file also describes how to compile these C codes by the mex compiler.

This program uses three external parameters, which are listed below. Note that, depending on the classifier that you will use, there may exist other parameters.

  • minCircleRadius: The size of the smallest circular object (tissue component) to be located.

  • winSize: The size of a salient subregion cropped around a tissue component.

  • K: The number of clusters for quantizing the salient subregions.