Radiology Consolidation | Fast and Furious | Self-Learning AI

“We were kind of the picks and shovels of the AI goldrush.”

Flywheel president and CEO, James Olson, on the role of the company’s imaging data management and sharing platform in AI.


Imaging Wire Sponsors

  • Carestream – Focused on delivering innovation that is life changing – for patients, customers, employees, communities and other stakeholders.
  • Focused Ultrasound Foundation – Accelerating the development and adoption of focused ultrasound.
  • Medmo – Helping underinsured Americans save on medical scans by connecting them to imaging providers with unfilled schedule time.
  • Pocus Systems – A new Point of Care Ultrasound startup, combining a team of POCUS veterans with next-generation technology to disrupt the industry.

The Imaging Wire

Radiologist Practices Officially Consolidating
It’s not exactly news to most in the industry that US radiology practices are consolidating, but a new JACR study just revealed some pretty surprising details on the magnitude of this trend. Comparing radiologist practice database figures from 2014 and 2018, the study found that the number of radiology groups in the US fell by an astounding 14.6% (4,908 to 4,193) over the four years, driven in large part by a 21.2% decline in single-specialty practices (2,812 to 2,216) and a clear (and typical) shift towards larger practices. Here’s the breakdown from a radiologist workforce share perspective:

  • Radiologists in 1 or 2 member practices fell from 3.2% to 2.1%
  • Radiologists in 3 to 9 member practices fell from 10.2% to 6.7%
  • Radiologists in 10 to 24 member practices fell from 18.2% to 14.1%
  • Radiologists in 25 to 49 member practices fell from 16.6% to 15.1%
  • Radiologists in 50 to 99 member practices fell from 13.3% to 11.5%.
  • Radiologists in 100 to 499 member practices grew from 15.7% to 21.8%
  • Radiologists in ≥500 member practices grew from 22.9% to 28.7%

Due in large part to ongoing practice acquisitions, the share of radiology practices with 100 to 499 members jumped from 7.6% to 10.2% and the share of groups with 500 or more members went all the way from 2.5% to 4.1%. Meanwhile, the share of one or two member practices fell from 26.9% to 22.8%.

Those are really big four-year shifts for a 120-year old industry and will likely have a major influence on care, costs, and compensation, not to mention the role of the partner career path. However, it’s important to remember that every industry in the world is somewhere on its path towards consolidation, so you might as well position your business or career strategy to take advantage of the opportunities that consolidation provides.

Fast and Furious
Fast Company triggered the radiologist and AI communities this week with an article about “free, open-source software (that) can diagnose us as well as doctors.” The not-for-medical use tool, called Chester the AI Radiology Assistant, was developed by a team of well-intentioned Mila Quebec researchers by combining a locally-operated AI algorithm based on a NIH chest X-ray dataset with a browser-based interface. With Chester, users can upload their own chest X-rays and “diagnose 14 diseases with 80% accuracy . . . about as well as a real radiologist,” while keeping patient data private (because it’s local). Chester isn’t intended to replace doctors, but its developers do suggest that future iterations (that will still be free) may serve as an alternative to commercial AI solutions.

The radiology community had good reason to be upset about this article from a scientific, clinical, economic, and journalistic perspective, especially given the hype around AI and physicians’ well-established distaste for self-diagnoses. That said, this type of coverage should be expected as the topic of medical AI migrates deeper into the sensational world of mainstream media, and I bet Fast Company would argue that this article got way more traction than a straightforward approach would have.

Carestream’s DR Focus
Carestream and AuntMinnie.com provided some new insights into Carestream’s DR-focused future following Philips’ acquisition of its healthcare IT division last month. Carestream primarily emphasized how the acquisition will allow it to increase R&D funding for its “core” digital radiography business, which in the near term means:

  • A focus on its Nano mobile X-ray system
  • The launch of a new low-dose 25×30-inch DRX detector
  • Creating a center of excellence for its value DR systems
  • Building upon its high-priority OnSight conebeam CT
  • Adding a new tomosynthesis application for its DRX-Evolution Plus x-ray room

However, it doesn’t appear to mean a major modality expansion is on the way, as Carestream clearly positioned its ultrasound line as internationally focused and only mentioned traditional CT and MRI when discussing DR’s advantages over the “advanced” modalities.

This is a helpful message for any company to broadcast following an acquisition of this magnitude, painting a clear picture that Carestream is DR-focused and has reasons to be “really excited about (its) future.”

The Self-Learning AI Debate
Speaking of AI news making its way into mainstream media, NPR’s All Things Considered introduced the debate surrounding self-taught AI to its ~3 million daily listeners. The story detailed self-taught “black box” algorithms’ migration to clinical settings and their potential to solve serious health problems, but also discussed issues that can arise when an algorithm is learning on its own (e.g. they sometimes “cheat” to get their findings). This made some healthcare AI-watchers to call for the development of more transparent and interpretable solutions or the use of “explanation model” algorithms to supplement black box AI. However, the story largely suggests that it’s more important for clinicians to know which solutions to use and when to use them, while understanding how an algorithm arrived at a diagnosis is less important. After all, “physicians use things that they don’t understand how they work all the time.”


The Wire

  • It didn’t take long after the NPR story for the FDA to announce plans to roll-out a new regulatory framework for self-learning AI algorithms, which is probably a good idea given the unique factors involved with regulating a product that is continually evolving. The FDA’s first step towards regulating these advanced AI solutions is a new discussion paper drafted this week, outlining the areas that the FDA may review (e.g. performance, modification plans, and ability to manage modifications), and then applying feedback to a future draft guidance.

  • Imaging informatics company, Flywheel Exchange, announced its completion of a $6 million Series A funding round that it will use to increase its sales and marketing efforts and potentially fund its expansion from academic research to the hospital and pharmaceutical markets. Flywheel is planning a later $5 million to $10 million convertible note round to fully expand to hospitals and pharma.

  • The CT Colonography Screening for Colorectal Cancer Act, which would require Medicare to cover CT Colonography for colorectal cancer screening, was introduced to the US House of Representatives last week with significant support from radiology and healthcare industry groups. In fact, industry groups have placed significant focus on CTC coverage for quite some time, stating that it would raise screening rates, reduce costs, and benefit marginalized groups who are often under-screened for colon cancer.


The Resource Wire

This is sponsored content.

  • Focus Ultrasound Foundation scientists published a study validating the performance of their Kranion software, intended to allow clinicians to plan and visualize hypothetical transcranial focused ultrasound procedures. Using MRI and CT images from 28 patients, the study found that Kranion was effective at predicting average temperature rise during ablation.

  • Did you know that imaging patients are most likely to no-show for their procedures on Mondays and Saturdays? By partnering with Medmo, imaging centers can keep their schedules full, despite the inevitable Monday no-shows.

  • This Innovatus Imaging blog highlights the company’s Ultrasound School program, designed to train students on managing imaging device inventory, repair methodologies, and device maintenance.

Leave a Reply

Basic HTML is allowed. Your email address will not be published.

Subscribe to this comment feed via RSS

Get every issue of The Imaging Wire

Delivered right to your inbox. We will never sell or share your information.