Abstract
Given the importance of gland morphology in grading prostate cancer (PCa), automatically differentiating between epithelium and other tissues is an important prerequisite for the development of automated methods for detecting PCa. We propose a new deep learning method to segment epithelial tissue in digitised hematoxylin and eosin (H&E) stained prostatectomy slides using immunohistochemistry (IHC) as reference standard. We used IHC to create a precise and objective ground truth compared to manual outlining on H&E slides, especially in areas with high-grade PCa. 102 tissue sections were stained with H&E and subsequently restained with P63 and CK8/18 IHC markers to highlight epithelial structures. Afterwards each pair was co-registered. First, we trained a U-Net to segment epithelial structures in IHC using a subset of the IHC slides that were preprocessed with color deconvolution. Second, this network was applied to the remaining slides to create the reference standard used to train a second U-Net on H&E. Our system accurately segmented both intact glands and individual tumour epithelial cells. The generalisation capacity of our system is shown using an independent external dataset from a different centre. We envision this segmentation as the first part of a fully automated prostate cancer grading pipeline.
Highlights
With 1.1 million new diagnoses every year, prostate cancer (PCa) is the most common cancer in men in developed countries[1]
The network output was compared with the ground truth: color deconvolution masks generated from the IHC slides with manual corrections
As the hematoxylin and eosin (H&E) network was trained on the output of the IHC network we considered the IHC performance as an upper bound for the performance of the H&E network
Summary
With 1.1 million new diagnoses every year, prostate cancer (PCa) is the most common cancer in men in developed countries[1]. The histological grade in PCa is formally defined in the Gleason grading system[3], and is a powerful prognostic marker It is determined by pathologists on hematoxylin and eosin (H&E) stained tissue specimens. The identification and grading of prostate cancer can be time consuming and tedious for pathologists, as all individual cancer foci within a surgical specimen or biopsy have to be analysed. Deep learning methods that try to detect or grade cancer from scanned tissue slides are typically trained using a set of annotated regions as the reference standard. As these algorithms learn from training data, the quality of the output is directly linked to the quality of the training samples. Tumor annotations made by pathologists are often coarse and contain large amounts of non-relevant tissue which adds noise to the reference standard and, subsequently, limits the potential of deep learning methods
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
