Impact of rescanning and normalization on convolutional neural network performance in multi-center, whole-slide classification of prostate cancer

Zaneta Swiderska-Chadaj,Geert Litjens,Jeroen Van Der Laak,Dirk Vossen,Lionel Blanchet,Thomas De Bel,Alexi Baidoshvili

doi:10.1038/s41598-020-71420-0

Abstract

Algorithms can improve the objectivity and efficiency of histopathologic slide analysis. In this paper, we investigated the impact of scanning systems (scanners) and cycle-GAN-based normalization on algorithm performance, by comparing different deep learning models to automatically detect prostate cancer in whole-slide images. Specifically, we compare U-Net, DenseNet and EfficientNet. Models were developed on a multi-center cohort with 582 WSIs and subsequently evaluated on two independent test sets including 85 and 50 WSIs, respectively, to show the robustness of the proposed method to differing staining protocols and scanner types. We also investigated the application of normalization as a pre-processing step by two techniques, the whole-slide image color standardizer (WSICS) algorithm, and a cycle-GAN based method. For the two independent datasets we obtained an AUC of 0.92 and 0.83 respectively. After rescanning the AUC improves to 0.91/0.88 and after style normalization to 0.98/0.97. In the future our algorithm could be used to automatically pre-screen prostate biopsies to alleviate the workload of pathologists.

Highlights

Algorithms can improve the objectivity and efficiency of histopathologic slide analysis
This paper has four main contributions: (I) we developed and compared different deep learning approaches that address prostate cancer detection at whole-slide image-level based on a multicenter dataset, (II) the proposed method was evaluated based on two independent datasets of 85 and 50 wholeslide images digitized on scanners from two vendors and from a medical center not included in the development set, (III) we took into account the influence of scanner variability on a deep learning classification results, and (IV) we investigated the influence of color and style normalization on classification results
The total development set consisted of 582 whole slide images (WSIs). 486 WSIs were used in a three-fold cross-validation procedure for network training, and 96 WSIs were kept separate to optimize post-processing hyperparameters

Summary

Introduction

Algorithms can improve the objectivity and efficiency of histopathologic slide analysis. We investigated the impact of scanning systems (scanners) and cycle-GAN-based normalization on algorithm performance, by comparing different deep learning models to automatically detect prostate cancer in whole-slide images. During the prostate biopsy procedure, 6–12 core samples are taken from a patient[3] resulting worldwide in more than 15 million specimens annually, which is expected to increase further with the aging of the population. All these specimens have to be evaluated by pathologists. We compare two different normalization approache (color and style normalization) and investigated their impact

Objectives

Methods

Results

Discussion

Conclusion