We know from past studies that, how important pathology is for breast cancer. It’s not just one disease, There’s lots of different types of breast cancer, but we’re still trying to really understand what that different types look like. And so, to really get to utilize machine learning tools to help us accelerate our understanding of the underlying
patterns within the breast tissue has been really efficient. We’re able to look at 1700 cases with tissue samples, apply that to over 20 years of data. And we were able to really do it in three months time, which would probably would have taken years
for a team of pathologists to do. Yeah and those images are 5 to 10 gigs a piece. Right? So if you multiply that out by 1,700, it’s massive. Yeah. We apply multiple layers of algorithms to basically do this unsupervised clustering. So we say, “Hey machine, can you go group all of these tiles that we have
in a way that makes sense to you.” The great thing is, is that some of those clusters
were actually things that we would expect it to find. And then some of the clusters were things
that we can’t describe at all because that’s the benefit of the unsupervised approach. And so, when we were considering cloud options, I think that Google’s commitment to privacy and being compliant with HIPPA
and other federal regulations was key to the selection. One of the things that I love about this project is it provides an example of how the American Cancer Society can partner with socially conscious
corporations like Slalom and Google and work together to make cancer research advances. All of us have been impacted by cancer in some way and so being able to even put a tiny little dent in that is, it’s just awesome.