Dr. Recht at RSNA 2018

Dr. Michael Recht urges tech, data and AI innovation to rescue imaging at RSNA

November 26, 2018
by John W. Mitchell, Senior Correspondent
For Dr. Michael Recht, a key “aha!” professional moment came when he was online planning a Disney vacation for his youngest daughter. A pop-up invite from a Disney vacation planner proved very helpful in making decisions about their upcoming visit, and in doing so it raised a question:

Why can’t we do that in radiology?

Recht, who gave the annual RSNA Oration in Diagnostic Radiology at the opening session in the nearly filled to capacity Arie Crown Theater at McCormick Center in Chicago, spends a lot of time puzzling over how to make imaging more useful to both referring physicians and patients. Using first-person insights from his role as chair of the radiology department at NYU Langone Health (NYULH), he dazzled the audience with the hard metrics of reinvention to improve service to both groups.

“Data needs to be in real time ... not two or three weeks later when I forget why I asked the question [concerning the data],” he said in his remarks. “Data needs to be customizable ... across departments ... and actionable.”

Innovative use of data, he argued, is the antidote to radiologist burnout, lower reimbursement, and increasing workloads. To that end, he detailed unique medical imaging delivery innovations rolled out at NYULH.

In a return to the days of live rounds with referring physicians, he and his colleagues devised an on-screen, live chat feature with an on-duty virtual radiologist. Through this portal, both the referring physicians and the radiologist were given control over the computer mouse to highlight any section of an image to aid in diagnostics and treatment.

At first, Recht said, some of his colleagues protested that such “virtual consults” would be an interruption to their day. But he convinced them that it was a worthwhile effort, quoting Dr. Deepak Kaura from a 2016 JACR article: “It is a privilege to be called at any time; it means that we are valued and that our opinions matter in the care of patients.”

Dr. Michael Recht
Recht also detailed how the department was able to reduce MR turnaround times to two minutes and 12 seconds – a decrease of five minutes and 28 seconds. The MR staff achieved this by reconfiguring throughput with several changes.

This included using dockable gurneys, adding technicians in the suite and building separate entrance and exit doors to eliminate congestion in and out of the MR suite. Ultimately creating the ability to scan three additional patients a day, increasing annual revenue by $390,000 per machine across the health system's 40 MR units.

This has heightened patient satisfaction as they are on the exam table within 15 minutes of their appointment time in 82 percent of cases with total patient time spent in the department at less than 90 minutes. And, Recht said, the additional revenue helped justify the institution’s research and extra staffing costs to achieve the metric.

Recht concluded with a discussion of the evolving role of AI in research at NYULH. He and his colleagues are progressing on a goal to acquire an MR scan ten times faster through data compression and reconfiguration through a machine learning neural algorithm. He presented several images of the standard and neural AI images that he challenged the audience to detect the difference between – there were no takers.

He also announced the availability of an open MR data source file composed of their MR speed research. And, he said, there will soon be an announcement of a “significant” prize for any clinical party that can advance NYLUH’s current data MR capture from a times-four factor to a times-ten factor.

“Our goal is to replace radiological exams with MR exams,” said Recht.