We are interested in developing an AI based scoring method to score our colitis samples. We do have necessary samples but need help in programming.
I'm not overly familiar with histology scoring but if you have a histology slice you could identify certain features (such as a lesion) if you had enough examples. If you are using Matlab and wish to use AI they have an image labelling tool (https://au.mathworks.com/help/vision/image-labeling.html) which feeds nicely into their different AI tools for image analysis. Additionally you could look at image segmentation methods to identify these features.
Once these features have been identified you could, as the previous answer suggested, then use this to train an algorithm to assign a score based on the number of features identified and the quality of these features.
Hope that helps
How many images to do you have?
what is your goal (detect colitis versus non-colitis?)
I can help you with this, please send me a private message.
Thanks!
What i get from asking a question here is to find a way to use AI instead of Macroscopic analysis where conventionally in Colitis (H&E) stains images mostly people look at Neutrophil Infiltration, crypt damage and its shape, size of mucosa and submucosa as well as distribution with abundance of Goblet cells. and then compare in different groups of interest. I was waiting for some AI tools and suggestions; instead of same macroscopic scoring sytem which mostly blind observer is required. am i wrong or right?
Thank you for your answers.
Weilei Mu, yes that is exactly what we thought to do. We are now at that stage where we have collected a considerable number of samples that we have all scored using annotations in an histologic image program. Now we somehow need to transfer the information into the algorithm .
Mark Gardner, Thanks for your suggestion, we will have a look into it. We do have an imaging program that we used, but I am not sure whether it is possible to use the image format of that program.
Ahmad Ud Din, absolutely right. We have developed a score with reduced features which is easy to score using digital slide scanning and image analysis. The score is validated by comparing it to clinical measures, now we want to go the next step for automated analysis.
- Badges
- Science topic
- Similar topics
- Computer Science
- Images