Welcome back. Researchers from Harvard Medical School have developed an artificial intelligence model capable of predicting colorectal cancer, patient survival, and what the optimal therapy might be for them to more accurately predict it than human pathologists. It is yet another breakthrough in the field of cancer care, and one that holds immense promise for patients and health care professionals alike. Earlier, I spoke with the head of the U Lab at Harvard Medical School that developed the technology, Dr. Kun Yu. Take a look.
Dr. Yu, thank you for being here. Thank you for having me. I'm fascinated by this. Walk us through how the tool actually works.
Yeah, so our tool, we named it MOMA, multi-omics, multi-cohort assessment, is a novel artificial intelligence tool where we trained the tool using a bunch of historical pathology data, that is the samples we collected for diagnosing cancer patients, and we leveraged those same pathological samples to predict which patient would likely respond to the current treatments.
So it's not that it can detect early colon cancer, but rather it is able to identify the chances of survival in someone who's already been diagnosed. Do I understand that correctly?
Yeah, in our current work, we focus on people who have already been diagnosed of colorectal cancer, and we try to predict their actual survival and response outcomes using their pathological data. And related to the early diagnosis and the cancer detection, there are a few prior works, including those from our labs working on that part. So how does it compare then to traditional methods of predicting colon cancer survival and best treatment plans?
Yeah, that's a great question. So currently, we do have a few clinical parameters that clinicians use to predict which patient would likely need more aggressive form of treatment. But unfortunately, those parameters can only stratify some major groups without actually pinpoint the personalized survival outcomes. And our tool is trying to bridge this gap by providing personalized clinical outcome prediction for all patients.
How confident are you in the reliability and accuracy of this model? Yeah, that's a very important bit for novel AI approach because we know many of these AI models are very complicated and they may have differential performance among different populations. And that is actually one of the key point in our publication as well. We validated our approach in multiple cancer patient populations where we successfully predict
the early stage colorectal cancer survival as well as the late stage, the more advanced form of colorectal cancer among more than three different patient populations. Dr. Yu, you talk about how the tool is used to identify the best treatment plan, but does it take into account personal preferences around quality of life and side effects, for example?
Currently, we are only looking at the response to the treatment, that is, after the standard treatment is applied, how long can the patient survive without cancer? And also, how long can the patient survive overall with or without cancers? And that's usually the main defining factor for patient's quality of life.
And we are currently working on developing a more advanced form of AI approach that can also consider, as you mentioned, the patient preference as well as their availability and access to treatments. Do you hope that this model will eventually evolve to be able to detect early colon cancer?
Yeah, totally. So in our current academic paper, we specifically focus on the patients who have been diagnosed with colorectal cancer. But we believe not only it can be extended into early stage colorectal cancer, but even all kinds of cancers will be able to benefit from this new form of technology.
And finally, I'm curious what your thoughts are on the potential for AI to replace doctors. It's a big point of discussion as AI continues to evolve, particularly in the health industry. What's your thoughts on that? Yeah, so we believe at the end of the day, it's not AI will take over the world and replace all the doctors. It's more like the doctors who know how to use AI will replace those who don't.
All right, well, we'll have to leave it there. Dr. Yu, thank you very much for your time. Thank you. We'll continue the discussion with the panel on the other side. Don't go away.