Automated diagnosis of 7 canine skin tumors using machine learning on H&E-stained whole slide images

Abstract

Microscopic evaluation of hematoxylin and eosin-stained slides is still the diagnostic gold standard for a variety of diseases, including neoplasms. Nevertheless, intra- and interrater variability are well documented among pathologists. So far, computer assistance via automated image analysis has shown potential to support pathologists in improving accuracy and reproducibility of quantitative tasks. In this proof of principle study, we describe a machine-learning-based algorithm for the automated diagnosis of 7 of the most common canine skin tumors: trichoblastoma, squamous cell carcinoma, peripheral nerve sheath tumor, melanoma, histiocytoma, mast cell tumor, and plasmacytoma. We selected, digitized, and annotated 350 hematoxylin and eosin-stained slides (50 per tumor type) to create a database divided into training, n = 245 whole-slide images (WSIs), validation ( n = 35 WSIs), and test sets ( n = 70 WSIs). Full annotations included the 7 tumor classes and 6 normal skin structures. The data set was used to train a convolutional neural network (CNN) for the automatic segmentation of tumor and nontumor classes. Subsequently, the detected tumor regions were classified patch-wise into 1 of the 7 tumor classes. A majority of patches-approach led to a tumor classification accuracy of the network on the slide-level of 95% (133/140 WSIs), with a patch-level precision of 85%. The same 140 WSIs were provided to 6 experienced pathologists for diagnosis, who achieved a similar slide-level accuracy of 98% (137/140 correct majority votes). Our results highlight the feasibility of artificial intelligence-based methods as a support tool in diagnostic oncologic pathology with future applications in other species and tumor types.

Automated Nuclear Morphometry: A Deep Learning Approach for Prognostication in Canine Pulmonary Carcinoma to Enhance Reproducibility

“…Deep learning-based tools, like nuclear morphometry algorithms, have been increasingly investigated for numerous diagnostic/prognostic tasks in veterinary and human pathology in the detection [ 47 , 48 ], classification [ 24 , 49 , 50 , 51 ], and grading of tumors [ 52 , 53 ]. These studies have proposed different ways for algorithms to be implemented into the diagnostic workflow (fully automated and computer-assisted diagnosis).…”

“…The current approach for assessing nuclear pleomorphism in cPC, following the methods of the 1997 grading system [ 7 ], is the categorical estimation into three classes based on the variation in nuclear size (anisokaryosis) and shape irregularity. Nuclear pleomorphism estimates by pathologists are considered to be limited by low reproducibility, as has been shown for canine mammary carcinoma [ 22 , 23 ] and canine cutaneous mast cell tumors [ 24 ] and is suspected for cPC.…”

Review of Molecular Technologies for Investigating Canine Cancer

“…However, more recently, genotype–phenotype correlation investigations have been performed as pan-cancer studies wherein an extensive range of genetic alterations are considered [ 133 , 134 , 143 ]. In veterinary oncology, investigations of deep learning-based tumour histology are starting to be performed on canine tumours, wherein artificial intelligence (AI)-based methods are using digitized whole-slide images (WSIs) to make pathological diagnoses, such as diagnosing canine skin tumours (of seven different types) [ 144 , 145 ] and canine osteosarcoma (where the AI program was able to identify specific histologic subtypes that may have prognostic value) [ 146 ]. However, to date, there has only been one genotype–phenotype correlation study in canines; it was recently shown that deep learning using a commercial AI histology software could detect the BRAF p.V595E mutation on HE-stained canine bladder cancer tissue sections ( Figure 9 ), with a sensitivity of 89% [ 147 ].…”