PathologAI Initiative
Initiative for an effective and accurate framework to analyze histopathological data from animal studies to advance digital pathology in preclinical application
![PathologAI Logo](https://public4.pagefreezer.com:443/content/FDA/28-09-2023T11:25/https://www.fda.gov/files/styles/main_image_medium/public/PathologAI.png?itok=Jm19c5VA)
Objective: To develop an effective and accurate framework for the analysis of histopathological data from animal studies to advance digital pathology in preclinical application.
Introduction: Preclinical pathology is a central component in toxicity evaluation and regulatory assessment with animal studies. However, it is a laborious process and requires extensive training. Nevertheless, discrepancies among pathologists still exist. Meanwhile, the rapid advancement in image analysis with AI has been extensively evaluated in clinical application with encouraging results. In this initiative, we are developing an AI framework for animal-pathology data which could facilitate the identification of lesion type, location, and severity in hematoxylin and eosin-stained slides.
Approaches: PathologAI is a multistage deep-learning framework for digital pathology analysis of whole slide images of preclinical animal studies. The framework includes specific functions to identify lesion type, severity, and location. The initial application of the proposed PathologAI framework in necrotic-lesion (prematurely dead tissue cells) identification is encouraging and the pinpointed lesion locations are consistent with the manual examinations by pathologists.
Potential Impact: The proposed PathologAI may quickly and accurately identify and quantify specific lesion types. It may also quantitatively evaluate morphologically relevant regions of interest to facilitate pathologists’ reviews in the preclinical setting.
References:
- A Weakly Supervised Deep Learning Framework for Whole Slide Classification to Facilitate Digital Pathology in Animal Study.
Bussola N., Xu J., Wu L., Gorini L., Zhang Y., Furlanello C., and Tong W.
Chemical Research in Toxicology. 2023, 36(8):1321-1331. doi.org/10.1021/acs.chemrestox.3c00058. Epub 2023 Aug 4. PMID: 37540590; PMCID: PMC10445282.