Inflammatory Bowel Disease (IBD) is a debilitating disease for which no cure exists. Murine models are widely used to study IBD, amongst them, the DSS-colitis model is a rapid and reliable one. However, the scoring of histopathological lesions is challenging.
We interviewed Dr. Agathe Bedard and Dr. Aleksandra Zuraw, veterinary pathologists at Charles River Laboratories, on their recently published work in developing a deep learning AI model with Aiforia's AI development tool, Aiforia® Create, to increase the accuracy and consistency of assessing DSS-colitis models, enabling the evaluation of different drug candidates for IBD.
"We used Aiforia for Dextran Sulfate Sodium (DSS) – induced colitis model evaluation in mice. In this model, the mice develop colon inflammation after administration of DSS, and the efficacy of different drug candidates targeting colitis can be tested. This model is very widely used for inflammatory bowel disease (IBD) research."
"Even though the DSS-colitis model is well established and reliable, the scoring of histopathological lesions is challenging. Several components need to be visually assessed and graded by a pathologist in order to calculate the final score. The disease induction is also segmental in nature, rendering scoring difficult.
Scoring is simpler in extreme cases (not affected vs severely affected animals), but in the middle of the spectrum, it is difficult for a pathologist to perceive subtle changes and evaluate how well exactly the candidate drug is working. In addition, if different pathologists evaluate different studies, the confounding factor of inter-pathologist variability starts playing a role in the interpretation. This reality makes the comparison across studies even more difficult.
Finally, the data generated with semi-quantitative scoring requires advanced statistical methods. We hypothesized, that an artificial intelligence model could be developed to recognize the relevant components of the DSS-induced colitis and that it could perform consistently within and across studies and help generate quantitative data better suited for routine statistical analysis."
"Aiforia, being a cloud-based platform, gave us the possibility to work on the same project independently on our own time, and there was no need to purchase or use dedicated hardware.
The platform could be accessed from our computers. At the beginning of the project, it was a commodity, something we liked very much, but later, it would be possible for us to organize our work differently.
Later, when Aleksandra transferred to a different CRL site, and the COVID-19 pandemic forced us to work from home, it became an indispensable feature of the software, enabling us to continue working on the project remotely."
"Starting the project was relatively easy. A standard level of computer literacy was enough to set up the project and get going. It was empowering that even without previous experience in deep learning model development, we were able to advance our project and see the first promising results. With the Aiforia team’s assistance, we learned the software better and are now able to work on future projects independently."
"The biggest benefit is the remote access to the software. Charles River Laboratories is a multinational, multisite company with over 150 pathologists. It is great to have a tool that enables us to collaborate with anyone, regardless of their location."
"Another important benefit is the ease of use of the platform, which helps us bring more pathologists on board, initiate more projects, and leverage the power of artificial intelligence and deep learning to solve problems similar to those we have encountered in the course of our project."
Read the full publication here →
Discover more case studies: