Automated Counting of Cancer Cells by Ensembling Deep Features.

Liu Liu,Murakami Murakami,Junker Junker,Hu Hu

doi:10.3390/cells8091019

Liu Liu, Murakami Murakami + Show 2 more

Open Access

https://doi.org/10.3390/cells8091019

Copy DOI

Abstract

High-content and high-throughput digital microscopes have generated large image sets in biological experiments and clinical practice. Automatic image analysis techniques, such as cell counting, are in high demand. Here, cell counting was treated as a regression problem using image features (phenotypes) extracted by deep learning models. Three deep convolutional neural network models were developed to regress image features to their cell counts in an end-to-end way. Theoretically, ensembling imaging phenotypes should have better representative ability than a single type of imaging phenotype. We implemented this idea by integrating two types of imaging phenotypes (dot density map and foreground mask) extracted by two autoencoders and regressing the ensembled imaging phenotypes to cell counts afterwards. Two publicly available datasets with synthetic microscopic images were used to train and test the proposed models. Root mean square error, mean absolute error, mean absolute percent error, and Pearson correlation were applied to evaluate the models’ performance. The well-trained models were also applied to predict the cancer cell counts of real microscopic images acquired in a biological experiment to evaluate the roles of two colorectal-cancer-related genes. The proposed model by ensembling deep imaging features showed better performance in terms of smaller errors and larger correlations than those based on a single type of imaging feature. Overall, all models’ predictions showed a high correlation with the true cell counts. The ensembling-based model integrated high-level imaging phenotypes to improve the estimation of cell counts from high-content and high-throughput microscopic images.

Highlights

To study the molecular mechanisms of complex diseases like cancer, microscopic images can provide valuable information, but it is often necessary to carry out a series of molecular biology experiments on several conditions [1,2]
We developed an integrated end-to-end deep convolutional neural network (DCNN) model (ERDCNN) to regress microscopic image features to image-specific cell counts
The model integrated the Density Map Regression DCNN Model (DRDCNN) model, which had a U-net as feature extractor, and the weights were initialized from a previous study, and the Foreground-mask-based regression DCNN (FRDCNN) model, which had a VGG-style autoencoder as feature extractor

Summary

Introduction

To study the molecular mechanisms of complex diseases like cancer, microscopic images can provide valuable information, but it is often necessary to carry out a series of molecular biology experiments on several conditions [1,2]. The images from the experiments have been evaluated manually. It was time-consuming and needed a lot of human effort and expertise. The number of cells in a microscopic image can be used as a measurement to be compared among different groups. We can evaluate the treatment effect of a cancer drug at different doses by comparing the microscope-based cancer cell counts under the specified conditions [2]. The experimental group with the smallest cell count number in its microscopic images may be treated as

Methods

Results

Conclusion