Using Deep Learning Artificial Intelligence Algorithms to Verify N-Nitroso-N-Methylurea and Urethane Positive Control Proliferative Changes in Tg-RasH2 Mouse Carcinogenicity Studies

Abstract

In Tg-rasH2 carcinogenicity mouse models, a positive control group is treated with a carcinogen such as urethane or N-nitroso-N-methylurea to test study validity based on the presence of the expected proliferative lesions in the transgenic mice. We hypothesized that artificial intelligence–based deep learning (DL) could provide decision support for the toxicologic pathologist by screening for the proliferative changes, verifying the expected pattern for the positive control groups. Whole slide images (WSIs) of the lungs, thymus, and stomach from positive control groups were used for supervised training of a convolutional neural network (CNN). A single pathologist annotated WSIs of normal and abnormal tissue regions for training the CNN-based supervised classifier using INHAND criteria. The algorithm was evaluated using a subset of tissue regions that were not used for training and then additional tissues were evaluated blindly by 2 independent pathologists. A binary output (proliferative classes present or not) from the pathologists was compared to that of the CNN classifier. The CNN model grouped proliferative lesion positive and negative animals at high concordance with the pathologists. This process simulated a workflow for review of these studies, whereby a DL algorithm could provide decision support for the pathologists in a nonclinical study.

Impact of Environmental and Pharmacologic Changes on the Upper Gastrointestinal Microbiome

“…Some organisms can produce Nnitroso compounds, some of which have been verified to be carcinogens. Several studies have hypothesized that shifts in the microbiome toward organisms capable of producing N-nitroso compounds increase the risk of gastric cancer [72][73][74][75].…”

Special Issue on Digital Pathology, Tissue Image Analysis, Artificial Intelligence, and Machine Learning: Approximation of the Effect of Novel Technologies on Toxicologic Pathology

“…These include proprietary in-house built solutions, such as AI models built to count ovarian follicles, 13 or to quantify changes within retinal layer morphology, 14 and detection of endothelial tip cells in the oxygen-induced retinopathy model, 15 as well the utilization of commercially available application for spermatogenic staging, 16 analysis of rodent cardiomyocytes, 17 to support scoring of dextran sulfate sodium-induced colitis mouse model histology, 18 enumeration of cynomolgus bone marrow histology, 19 quantitative evaluation of hepatocellular cell hypertrophy in rats, 20 quantitate cell proliferation via common immunohistochemical biomarkers, 21 and for verification of changes observed in the Tg-rasH2 mouse used in carcinogenicity studies. 22 A fluorescence-based image analysis use-case (commercial software) is provided by Wilson et al 23 As novel applications at the periphery of the breadand-butter imaging work of a toxicologic pathologist are continuously emerging, Rousselle et al introduce a digital 3D topographic microscopy technique called scanning optical microscopy to evaluate re-endothelialization of vascular lumen after endovascular procedures. 24 Due to a substantial knowledge gap between those who are using AI-based tools in pathology, and those who do not, it is becoming increasingly challenging to write and publish papers on the subject that both contain the needed technical details and appropriate terminology to enable reproducibility of scientific data generation, as well as are written in a language that is accessible to all Toxicologic Pathology readers.…”

Mini Review: The Last Mile—Opportunities and Challenges for Machine Learning in Digital Toxicologic Pathology

“…The ML needs to address well-defined, focused end points such as identification and characterization of specific changes in animal efficacy and safety models. 6 In our experience, the ML algorithms that appear most helpful in the workflow for a toxicological pathologic study evaluation are those that provide an automatic discrimination of normal and abnormal tissues ("abnormality detection"). Nonclinical safety studies present the pathologist with complex, multidomain data that drive the pathologist to make considerations on the safety of the investigated compound.…”

“…Three in particular include (1) classification across magnifications: For many classification tasks, the context needed by the computer to accurately identify a specific class is limited by training at a single magnification 6 ; (2) grading uncertainty: The classes of some microscopic changes (eg, lung hyperplasia vs adenoma) differ much more subtly than other classes, causing annotation and classification uncertainty 6 ; and (3) rare event measurements in toxicologic pathology; some classes may be rare, such that obtaining adequate amounts of training data for such classes is challenging. 6 To address the first challenge (classification across magnifications): Multimagnification analysis methods that resemble how pathologists analyze histologic slides using microscopes have already been developed and implemented in breast cancer diagnostics 20 and are being developed in the nonclinical space. 21 This multimaginfication approach combines the patch-level information with the information gathered from the context of larger fields of view at lower magnification and have shown improved performance in comparison to singlemagnification classifiers.…”

“…21 In the TgRasH2 mouse model work, stomach papillomas and hyperplasia were not well differentiated at the Â10 magnification, and when a lower training magnification was employed (Â2.5), the additional context resulted in high performance (F1 scores). 6 For the second challenge (grading uncertainty), ML algorithms could help standardize our individual grading and thresholds across a variety of studies or even across pathology groups and laboratories. The guidance an ML algorithm provides on grading or thresholds is based on probability, and although the algorithm may indicate a strong likelihood for a certain class (a specific grade or diagnosis), this prediction can still be highly uncertain, mislead the pathologist, and result in false-positive or false-negative predictions.…”