This is the first year that Oregon Health and Science University has participated in the InventOR Collegiate Challenge and their finalist team, SHIFT is coming in hot with an innovative disease detection imaging tool that puts expensive imaging technology within reach using artificial intelligence.
This invention holds the potential for changing the way doctors think about medical imaging. Read on to learn more about domain translation, the piles of money that could be saved, and why Team SHIFT is sending a big "thank you' to Grace Hopper.
What is your invention?
The product we are demonstrating is a tool that takes as input one type of image and from it generates another type of image. The input image is an H&E image, which are cheap to produce and widely used in pathology to diagnose disease. The output image is an immunofluorescence image, which are typically very informative yet very expensive to produce.
Our team has trained an artificial intelligence system to do this translation which eliminates the need for an actual immunofluorescence image.
We believe this work will accelerate the rate of image production by three orders of magnitude, while cutting the cost from thousands of dollars to dozens of cents.
Where did you get the inspiration for your invention idea?
The fundamental approach we’re taking is that of “domain translation”, which is a popular technique used in many settings. The novelty of our approach is to apply these techniques to medical imaging “domains” (that is, H&E and immunofluorescence).
Our research group recognizes the value in immunofluorescence imaging for understanding disease, and became interested in figuring out if there exists a cheaper means of generating these images.
As you prepare to pitch your invention to judges, what are you most excited about sharing?
We are excited to share the progress we’ve made upon a single idea and carrying it through the various stages of commercialization thinking. We believe our team is very well connected with cutting-edge research being done at OHSU to rethink the way we treat and diagnose cancer. This work has grown out of innovative clashes between biologists, technologists, and computationalists working together.
Who are your invention heroes?
Elon Musk for doing crazy things with highly explosive rocket fuel, Steve Jobs for looking so damn good when pitching new ideas, and Grace Hopper for transforming the way humans and computers interact.
Tell us about your team. How did you connect with one another and how do you work together?
Erik and Geoff are graduate students in Young Hwan’s lab. Erik and Young Hwan launched the project idea as part of Erik’s internship in the Computational Biology Program at OHSU. Geoff joined the lab later and has been working on similar artificial intelligence systems ever since.
What is the most important thing you want the judges to understand about your invention — what will set you apart from the competition?
We are excited to share a vision for a radically new way of thinking about the generation of medical images. Imagine if instead of needing to acquire images for each and every sample (which consumes valuable time, effort, and harmful chemical reagents), what if we teach a computer to leverage information contained in thousands of such images to produce a best guess of the result?
Our approach would save thousands of dollars and weeks of processing for each and every sample.
At some point, our system might be as good as or – considering the prevalence of technological artifacts in the actual staining process – better than the actual staining process. More data, and more thinking, are still required to answer these questions.