Histological stain evaluation for machine learning applications

Abstract

Aims:A methodology for quantitative comparison of histological stains based on their classification and clustering performance, which may facilitate the choice of histological stains for automatic pattern and image analysis.Background:Machine learning and image analysis are becoming increasingly important in pathology applications for automatic analysis of histological tissue samples. Pathologists rely on multiple, contrasting stains to analyze tissue samples, but histological stains are developed for visual analysis and are not always ideal for automatic analysis.Materials and Methods:Thirteen different histological stains were used to stain adjacent prostate tissue sections from radical prostatectomies. We evaluate the stains for both supervised and unsupervised classification of stain/tissue combinations. For supervised classification we measure the error rate of nonlinear support vector machines, and for unsupervised classification we use the Rand index and the F-measure to assess the clustering results of a Gaussian mixture model based on expectation–maximization. Finally, we investigate class separability measures based on scatter criteria.Results:A methodology for quantitative evaluation of histological stains in terms of their classification and clustering efficacy that aims at improving segmentation and color decomposition. We demonstrate that for a specific tissue type, certain stains perform consistently better than others according to objective error criteria.Conclusions:The choice of histological stain for automatic analysis must be based on its classification and clustering performance, which are indicators of the performance of automatic segmentation of tissue into morphological components, which in turn may be the basis for diagnosis.

Glandular Segmentation of Prostate Cancer: An Illustration of How the Choice of Histopathological Stain Is One Key to Success for Computational Pathology

“…The traditional prostate tissue stain, H&E, which dates from 1896 (Mayer, 1896; Lillie, 1965), does not allow automatic identification of the prostate glandular structure which is key in computer-aided malignancy grading. In Azar et al (2013), the authors demonstrate that H&E is not ideal for machine learning applications, but that other stains, such as PSR-Htx (Puchtler et al, 1973; Junqueira et al, 1979), perform better, chiefly because the stain of the stroma is distinct from the stain of the epithelium. PSR stains the connective tissue surrounding the glands red, allowing precise identification of the glandular borders, which is required for gland segmentation.…”

“…Pathologists rely on multiple, contrasting stains to analyze tissue samples, but histological stains are developed for analysis with a microscope and not for computational pathology applications. In Azar et al (2013), several different histological stains were evaluated for automatic classification of components in prostate tissue. The stains were tested with both supervised and unsupervised classification methods which showed that some stains consistently outperform others according to objective error criteria.…”

A Regulatory Science Initiative to Harmonize and Standardize Digital Pathology and Machine Learning Processes to Speed up Clinical Innovation to Patients

“…For example, the variability of a stain (e.g., hematoxylin and eosin across or within laboratories) may influence the performance of a downstream mutation prediction algorithm. [ 19 20 21 ] In this example, one may consider drawing an arbitrary boundary before the staining step; however, the fixation and processing method (e.g., formalin fixed, paraffin embedded) or even the tissue acquisition, handling, or image acquisition[ 22 ] may influence the performance of the predictor as well. Thus, for the purpose of the Alliance , we considered three descriptors for the solution.…”

Adriamycin‐induced nephropathy in rats: Functional and cellular effects characterized by MRI

“…ED1 labeled slides were scanned with SanScope (Aperio) at Â40 magnification. The whole surface of each kidney section was manually delineated and the color corresponding to the ED1 was extracted by threshold setting (17). The ED1-area within the delineated region was automatically calculated.…”