Deep learning for necrosis detection using canine perivascular wall tumour whole slide images

Abstract

Necrosis seen in histopathology Whole Slide Images is a major criterion that contributes towards scoring tumour grade which then determines treatment options. However conventional manual assessment suffers from inter-operator reproducibility impacting grading precision. To address this, automatic necrosis detection using AI may be used to assess necrosis for final scoring that contributes towards the final clinical grade. Using deep learning AI, we describe a novel approach for automating necrosis detection in Whole Slide Images, tested on a canine Soft Tissue Sarcoma (cSTS) data set consisting of canine Perivascular Wall Tumours (cPWTs). A patch-based deep learning approach was developed where different variations of training a DenseNet-161 Convolutional Neural Network architecture were investigated as well as a stacking ensemble. An optimised DenseNet-161 with post-processing produced a hold-out test F1-score of 0.708 demonstrating state-of-the-art performance. This represents a novel first-time automated necrosis detection method in the cSTS domain as well specifically in detecting necrosis in cPWTs demonstrating a significant step forward in reproducible and reliable necrosis assessment for improving the precision of tumour grading.

Investigating the potential of untrained convolutional layers and pruning in computational pathology

“…This F1-score was further improved to 0.754 when optimal decision thresholds predetermined on each fold's validation set were applied as seen in figure 2 and Table 3. This also increases our previously reported F1-score of 0.708 in Rai et al 13,15 where a different feature extraction set-up with an additional post-processing step was used.…”

“…In this paper we use necrosis detection 13,14,15 as an exemplar problem in computational pathology drawn from a canine Perivascular Wall Tumours (cPWT) data set. An external veterinary pathologist diagnosed Canine Soft Tissue Sarcoma (cSTS) histology slides obtained from the Department of Microbiology, Immunology and Pathology, Colorado State University.…”

Detection of Necrosis in Digitised Whole-Slide Images for Better Grading of Canine Soft-Tissue Sarcomas Using Machine-Learning

“…In this study, we present a detailed report describing the automatic assessment of WSIs for the detection and quantification of necrosis in cSTSs, providing further insight and analysis from our baseline approach as previously published [ 27 ]. The experiments presented in this study confirmed that DenseNet161 is able to recognise areas of necrosis with high accuracy (92.7%).…”

“…There were several motivations for this study. We previously published [ 27 ] the first report on the use of deep learning to detect cSTSs in haematoxylin and eosin (H&E)-stained whole slides. However, the study reported here builds on the initial study and focuses on grading.…”

Feline Soft Tissue Sarcomas: A Review of the Classification and Histological Grading, with Comparison to Human and Canine

“…Another issue is the subjective nature of the inflammation score—although the differentiation score it replaces in the Trojani scheme [ 32 ] is also subjective, it would be advantageous if all criteria within any grading system were objective, readily obtainable from routinely-stained haematoxylin and eosin sections and easy to reproduce, thereby reducing variability between pathologists and laboratories [ 48 , 49 ]. With the advent and increasing adoption of image analysis within veterinary pathology [ 50 ] it may be that artificial intelligence plays an important role in quantifying criteria such as inflammation, as well as other features such as the extent of necrosis [ 51 ], and mitotic counts [ 52 ] with increased accuracy.…”