button-radnet
12 December, 2018

Brainstorm: RadNet and its AI Partners Make Imaging Smarter

Dr John Crues RadNet AI Radiology

2018 marked the 8th year RadNet co-hosted a booth with eRAD at RSNA. As part of our commitment to thought leadership in imaging, we submitted feature articles for Radiology News Daily, a publication distributed at RSNA. We hope you enjoy this post about RadNet and how we are Leading Radiology Forward.

John Crues, M.D., Medical Director of RadNet, discusses how the company is leveraging artificial intelligence (AI) for clinical and business gains. RadNet is the largest owned outpatient imaging provider in the U.S., and its AI strategy focuses on real-world application of the technology across its 300+ centers.

What is AI exactly?

Normally, computers take data, process it, and display it to a radiologist. AI goes a step further; data is processed in a way that is similar to how the brain processes information.

The first AI researchers understood how neurons work and created neuron-like functions in software—firing and not firing just like the brain. Over the last seven years, the algorithms have become much more sophisticated and extremely accurate. They’re still one millionth the complexity of the human brain, but we’ve come a long way.

Will AI really change care delivery in imaging?

We believe it will be essential. Five years from now, in order to be a competitive radiologist in the U.S., you’re going to have to be using AI.

Most of the development now is in academic institutions, but RadNet has partnered with several researchers to create algorithms that are useful in practical application—in other words, for a business that has patients. This is one of RadNet’s strengths—strategic partnerships. We’re partnering with Harvard, Stanford, Massachusetts General, an AI company called Nulogix, and others on a variety of projects that will have clinical and operational impact.

In what way?

From a clinical standpoint, our partners at Mass General have an algorithm that can detect acute brain issues, such as a hemorrhage. Once AI detects that issue, it can alert the neuro-radiologist so that the read can be performed immediately. The study goes into SuperSTAT status. We’re implementing that across all centers.

In the near term, we see the biggest impact on our business operations. We’re implementing AI to improve the accuracy of our billing and collections.

Another project has to do with our managed care operations. RadNet has practiced utilization review for decades, and soon Medicare will require that review—clinical decision support—on all advanced studies for their patients. We’re developing AI algorithms to determine whether the ordered study is appropriate, so it’s a partial automation of CDS.

How far will it go in study interpretation?

I don’t anticipate that it will even partly replace a radiologist in the next decade. AI is a tool to support radiologists in efficiency and accuracy. But over the next five years, I believe it will progress so much that it will almost be malpractice for a radiologist not to use AI algorithms to assist them in their interpretations. We’ll deliver faster, better care for patients.

12 December, 2018