Bone marrow cellularity is the volume ratio of hematopoiesis (the body’s manufacturing of blood cells) and fat, and decreases with age. Many compounds can alter the cellularity of bone marrow making it a vital tissue for analysis by pathologists in pharmacologic safety studies. Quantitative data is vital for statistical evaluation in risk and toxicity assessments.
However, the important analytical information of these studies are often represented by images, which lack unambiguous and objective criteria for analysis as interpretation of these images is largely affected by the experience of the researcher. Specialists in toxicology are often also not trained in translating results into mathematical statistical models, while specialists in mathematical statistics and modelling lack knowledge on the specificity of risk assessment and methods in toxicology.
AI models provide a more objective quantification of bone marrow cellularity and increase the reliability of image analysis by evaluating a larger area and providing accurate cell counts, compared to a pathologist’s subjective severity score. Objective data is also statistically analyzable to evaluate against other study data and the nature of AI model generation for image analysis allows for continuous improvement of the AI model with new training data.
A research team at Charles River Laboratories (CRL) developed an AI model with Aiforia Create to evaluate whole slide images of macaque sternebrae to identify and enumerate bone marrow hematopoietic cells. The AI model was trained and able to differentiate the hematopoietic cells from the other sternebrae tissues and serves as an objective measure of bone marrow cellularity in tissue sections. Learn more about Mark Smith, PhD VMD, Veterinary Pathologist at CRL on his latest publication in our interview with him below and from this on-demand webinar.
Mark Smith: Bone marrow is a standard tissue for analysis in pharmacologic safety studies.
Xenobiotic compounds can alter the cellularity of bone marrow. Anatomic pathologists evaluate bone marrow in hematoxylin and eosin stained bone sections and diagnose increases or decreases in cellularity and assign a subjective severity score. We felt an Artificial Intelligence model might be able to determine cellularity in a more objective manner and serve as a screening tool to alert study pathologists to changes in bone marrow cellularity.
There is no current AI model that quantifies the number of cells in the bone marrow from a histologic section.
The Aiforia platform is cloud based and accessible without the need to install software and was fairly easy to learn.
After an hour or so of instruction, training the model was straight forward as was subsequent troubleshooting to improve the model. In short, a smooth start.
The model will provide useful data on bone marrow cellularity which can be used to screen slides and draw attention to potential lesions. A similar effort by a pathologist would be very time consuming and not practical.
Smith MA, et al. Toxicologic Pathology, 2021