There’s a huge potential for AI models to help in hematology. I think one area where perhaps it’s got easier traction to start with is going to be in the sort of image recognition field because that’s where AI has really come into its own in radiology, in dermatology, in ophthalmology, in those areas. And again, you know, we’ve got blood films, we’ve got bone marrow aspirates, we’ve got bone marrow trephines, and there’s already some work showing that, you know, you can do quite a good prediction there...
There’s a huge potential for AI models to help in hematology. I think one area where perhaps it’s got easier traction to start with is going to be in the sort of image recognition field because that’s where AI has really come into its own in radiology, in dermatology, in ophthalmology, in those areas. And again, you know, we’ve got blood films, we’ve got bone marrow aspirates, we’ve got bone marrow trephines, and there’s already some work showing that, you know, you can do quite a good prediction there. It could help with efficiency in terms of identifying high-risk areas in a sample. So I think that’s probably going to be the easiest area.
I think with other areas, in terms of maybe prediction or treatment selection, I think one thing I’ve learned is that actually our existing statistical models are actually pretty good. So things like your Cox proportional hazards ratios that we use all the time are actually pretty robust. And it’s not going to be, I think, until we get really good, well-curated data sets from real-world practice perhaps, or combining lots of clinical trial data, that we’re going to be able to have the cleanness of data to be able to run those models to start looking for patterns in, I don’t know, in toxicities or in responses to treatment. But in some ways, we’ve got the infrastructure to do it, it’s just having the stakeholder engagement so that we can start getting all these data out of their silos into big repositories that we can start to analyze.