Skip to content
BY 4.0 license Open Access Published by De Gruyter September 18, 2019

Using Deep Correlation Features to define the Meta Style of Cell Images for Classification

Simon Grützmacher, Ralf Kemkemer and Cristóbal Curio

Abstract

Digital light microscopy techniques are among the most widely used methods in cell biology and medical research. Despite that, the automated classification of objects such as cells or specific parts of tissues in images is difficult. We present an approach to classify confluent cell layers in microscopy images by learned deep correlation features using deep neural networks. These deep correlation features are generated through the use of gram-based correlation features and are input to a neural network for learning the correlation between them. In this work we wanted to prove if a representation of cell data based on this is suitable for its classification as has been done for artworks with respect to their artistic period. The method generates images that contain recognizable characteristics of a specific cell type, for example, the average size and the ordered pattern.

Published Online: 2019-09-18
Published in Print: 2019-09-01

© 2019 by Walter de Gruyter Berlin/Boston

This work is licensed under the Creative Commons Attribution 4.0 Public License.

Scroll Up Arrow