The Impact of Content Based Image Retrieval on Pathologists
I am Ul Balis, I am the Chair of the Strategic Advisory Board of Inspirata.
CBIR is Content Based Image Retrieval. It’s analogous to matching images with images or searching for images with images. For those individuals that have looked at Google Goggles, the idea that if you take a picture of something, you can submit that picture to Google and then receive a list of matching items and what they are. In the realm of pathology and specifically surgical pathology, our pathologists are interested in looking at the images, the histology of patients’ tumors and malignancies. This is a visual science largely based on experience and expertise and it’s not uncommon that types tumors will present themselves that have a morphology or pattern which is unusual that one has to ask the question, “what is this?” CBIR as an algorithm allows for the matching of an image itself. In other words, this is an image that I have not seen before or have not seen often, what does this image look like as compared to a library of images that we’ve kept and stored ahead of time. By using content based, that image by computer algorithms is matched with a library of images, that library of images, in this case former cancer cases is presented back to the pathologist along with the add that was stored at the time when those images were created which will be the diagnosis, presumably then other pieces of valuable information, the survival or clinical course of the patient. And this immediately provides a very powerful tool for both predictive analytics, prognostics and therapeutics, because by comparing the current case with the matching library of prior cases, then the pathologist can be informed of: A. What is this, and B. what is it’s biological potential; by this I mean is this going to recur, is this a good or bad malignancy, what is the survival. These are all things that can inform the patient and their oncologist of appropriate next steps. Right now, the repertoire of tools in pathology doesn’t have this, we can’t say, “I have an image. I want to use the image to go back to a library of images”….We look up things by text. CBIR is the enabling of matching images to images themselves and this was not possible before the advent of digital pathology. With digital pathology, we have the ability now to take all of the cases we see and digitize them and put them into electronic libraries. That is the opening capability to now consider CBIR. It’s a very, very powerful concept that enables an elevated level of practice to pathology.
Inspirata is transforming cancer diagnostics by leveraging the power of computer algorithms and big data analytics to match archived pathology images that are coupled with the corresponding patient treatments and outcomes to new images to help develop predictions about new cancer cases. To learn more visit https://www.inspirata.com/cidt/.