Dr. Kavoussi discusses machine learning and AI-based tools’ roles in kidney stone treatment

In this interview, Nicholas L. Kavoussi, MD, discusses his clinical practice and research interests pertaining to kidney stones. Kavoussi is an assistant professor of urology at Vanderbilt University Medical Center in Nashville, Tennessee.

Transcription:

Please provide an overview of your clinical practice and research interests.

My clinical practice is in minimally invasive treatment of benign prostate disease and kidney stone disease, mainly through natural orifice surgery. My research interest is in developing novel tools to improve endoscopic surgery of kidney stone disease.

What is the biggest unmet need in kidney stone treatment/prevention?

I think the biggest unmet need is that kidney stone disease is common, and after you have a kidney stone, recurrence is common. We don't have a great way of identifying specific risk factors, and adequately mitigating those risk factors to prevent recurrence events and stone formation. The hope is by doing some of the work here, where we're trying to improve our endoscopic surgical ability using novel tools, we can more accurately treat stones and prevent future stone recurrence episodes.

Please discuss your work in machine learning as it pertains to kidney stones.

We have research support from the NIH through a novel technology R21 grant. The intent of this project is to improve endoscopic visibility and navigation during kidney stone surgery. So we're building maps and tracking kidney stones in low visibility settings. The goal of this project is to be able to teach computers what we see when we operate. That way, the computer can know what surgery is supposed to look like and build these maps and track kidney stones while we operate to make us more accurate and track potential harmful fragments while we treat these kidney stones.

What do you think will be the biggest story related to kidney stones in the coming years?

I think these machine learning and artificial intelligence-based [tools] will impact the way we practice clinically, especially with chronic kidney stone disease. I think our biggest issue right now is we don't have the numbers and the datasets needed to really evaluate these technologies and how they might impact our lives and the lives of our patients. So I think the role of these technologies, though they're still in their infancy and have a long way to go, will be pretty impactful in terms of how we treat and take care of patients kidney stone.

Is there anything you would like to add?

I think the key to remember for a lot of what we're doing is that kidney stone disease is complicated. It's a really heterogenous patient population. There are a lot of reasons why stones form. There are a lot of reasons why they recur; we don't quite understand that. I think the way we work up stones now is very variable from provider to provider, and really needs to fit patient's needs rather than our clinical workflow. My hope is that these machine learning-based tools and computer vision models will really allow for patient-directed, specific care. And in terms of how we treat the stone surgically and the tools we're building, this is really 1 facet of stone care. And I think really, it's important to consider all different factors that contribute to stone disease and recurrence when helping these patients.

This transcript was edited for clarity.