• There are no suggestions because the search field is empty.

AI in toxicologic pathology: an overview

Toxicologic pathology, the study of drug effects on animal models of human diseases, has benefited greatly from the development of AI, particularly deep learning neural networks.
Written by Aiforia

Toxicologic Pathology

Toxicologic pathology, the study of chemicals’ adverse effects on animal models of human diseases, has a significant use in preclinical drug assessment. Toxicologic pathologists are crucial for drug discovery and chemical safety assessments due to their skills in informational interpretation from a broad range of disciplines. This is particularly relevant for the various applications pathological data can be integrated into, such as identifying therapeutic targets, evaluating drug efficacy and mechanisms of toxicity, and validating animal models.

The work of a toxicologic pathologist is often meticulous and repetitive. Identification of changes in numerous slides of similar tissue sections is crucial for the analysis of a toxin. However, this tedious work is a target for mistakes, including inter- and intra-observer bias, miscalculations, etc. These aspects are where digitization and artificial intelligence (AI)-supported image analysis have significant influence.

AI in Toxicologic Pathology Today

The potential applications of digitized and AI-assisted image analysis are extensive, for example decreasing error rate in characterization of microscopic lesions by toxicologic pathologists. Simply digitizing the slide allows an AI model to do in seconds what takes a person minutes. AI can sort normal from abnormal samples allowing toxicologic pathologists to spend more time evaluating altered phenotypes instead of normal tissues, reducing the weeks necessary for evaluation and/or peer review processes that substantially affect drug development timelines.

Microscopic review of 100s to 1000s of slides per study is labor intensive and time consuming, particularly with patient-dependent factors at play. Many normal background lesions, which need to be identified as such, are sex and age-dependent and further influenced by species and strain. Lesions can be compared across studies, species, and databases for educational, diagnostic, or research purposes. It will also become easier to track the progression of a complex lesion through drug trials.


Quantifying the number of cells in the bone marrow from a histologic section with the help of AI

Case study examples

  1. Dextran Sulfate Sodium (DSS) – induced colitis model evaluation in mice to test efficacy of different drug candidates targeting colitis and inflammatory bowel disease (IBD)
    Case study: AI model for DSS-induced colitis 
  2. Screening For Bone Marrow Cellularity Changes in Cynomolgus Macaques in Toxicology Safety Studies Using Artificial Intelligence Models
    CRL bone marrow infographic
  3. Using artificial intelligence to identify and categorize rodent ovarian follicles for reproductive toxicity assessment
    CRL ovarian follicle infographic 

Pathologist’s Assistant

Interpretive diagnoses are subjective evaluations that may differ even between well-trained individuals. Whole slide imaging (WSI) for digitization of cell or tissue samples allows for data to be shared online and worldwide collaboration. Incorporating AI into digital image analysis improves the pathology workflow further, enabling faster and more accurate analysis, including predicting novel drug targets and toxic effects of drugs. There is also a push to minimize animal use in research and testing with an increased focus on high-throughput screening assays. As automated image analysis of WSIs becomes more common in nonclinical toxicology studies, AI will follow suit.

Artificial intelligence is particularly good for completing repetitive, detailed tasks quickly and accurately, beyond the abilities of human pathologists. However, AI will not replace all human tasks, rather assist and augment human efficiency. Research has shown that a collaborative effort between pathologist and AI model is more effective than either one alone. Pathologists provide cognitive reasoning and an understanding of various fields that, although a model can compute and analyze, it is ultimately the pathologist’s responsibility to evaluate the findings. A combination of emerging technology and creativity of toxicologic pathologists are transforming the field from an analog past to a digital and efficient future.


Toxicologic Pathology, 2020

Pharma Manufacturing, 2020

Toxicologic Pathology, ScienceDirect

Haschek and Rousseaux's Handbook of Toxicologic Pathology, 2013