Data Augmentation for Morphological Analysis of Histopathological Images Using Deep Learning

Abstract

In this study, we introduce a data augmentation procedure for histopathology image classification. This is an extension to our previous research, in which we showed the possibility to apply deep learning for morphological analysis of tumour cells. The research problem considered, aimed to distinguish how many cells are located in a structure composed of overlapping cells. We proved that the calculation of the tumour cell number is possible with convolutional neural networks. In this research, we examined the possibility to generate synthetic training data set and to use it for the same purpose. The lack of large data sets is a critical problem in medical image classification and classical augmentation procedures are not sufficient. Therefore, we introduce completely new augmentation approach for histopathology images and we prove the possibility to apply it for a cell-counting problem.

Konrad Karanowski

Machine Learning Researcher

Machine Learning Researcher focused on computer vision, generative models and AI for science.