AI Not a Threat | Surprise Billing Targeted | AI for UG-RT

“Scientists crowed; radiologists cowered; ventures capitalized.”

Stanford University professor, Curtis P. Langlotz, on the imaging industry’s reactions to the emergence of radiology 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
  • Nuance – AI and cloud-powered technology solutions to help radiologists stay focused, move quickly, and work smarter
  • Pocus Systems – A new Point of Care Ultrasound startup, combining a team of POCUS veterans with next-generation genuine AI technology to disrupt the industry
  • Qure.ai – Making healthcare more accessible by applying deep learning to radiology imaging

The Imaging Wire

AI Won’t Steal Your Job
Stanford’s Curtis P. Langlotz reassured radiologists that the threat of AI stealing their jobs is “overblown” and although AI will “profoundly change how radiologists practice,” these changes will most likely be “in a direction that pleases” them. This is far from the first reassurance that AI won’t steal radiologists’ jobs, and it probably won’t be enough to soothe some in the specialty, but Langlotz does provide solid evidence to support his point:

  • Remember CAD – Much like mammography CAD systems, which enjoyed early buzz and largely didn’t live up to expectations, Langlotz warns that the first wave of AI algorithms might not be accurate or generalizable enough to have a major clinical impact.
  • Remember MRI – Despite predictions that MRI systems would bring the demise of radiology, MR imaging proved to be very reliant on radiologists. The same is happening with AI as radiologists are becoming relied upon to “recognize AI’s shortcomings and capitalize on its strengths.”
  • There are A Lot of Diseases – AI may be valuable for identifying a single disease or a small set of diseases, but radiologists are relied on to diagnose thousands of diseases, including uncommon ones.
  • Tech Doesn’t Steal Jobs – There’s a long history of the media and other groups predicting that technology will steal jobs, but there’s little evidence of this actually happening.
  • Autopilot Still Takes Work – Much like airplane pilots during autopilot mode, AI will very likely handle the repetitive and easier work (e.g. finding breast calcifications, measuring and plotting liver metastases), while radiologists leverage their human qualities (e.g. judgement, creativity, empathy) as they “interpret images in partnership with AI algorithms and patients.”

The debate over AI’s impact is sure to continue, but this may be the last time it’s covered this directly by Langlotz, who ended his article by saying “Will AI replace radiologists? is the wrong question. The right answer is: Radiologists who use AI will replace radiologists who don’t.”

Congress Targets Surprise Billing, Twice
One week after President Trump called for Congress to address surprise medical billing, a pair of bipartisan groups from the U.S. House and Senate each proposed legislation to curb the unpopular practice. Both bills target the unexpected charges that occur when patients go to an in-network provider but are treated by an out-of-network clinician (often radiologists or anesthesiologists) or when a patient goes to an out-of-network provider (often medical emergencies). However, the two bills are taking very different approaches:

  • Senate – The new Senate bill proposes an arbitration program to handle surprise billing disputes, an approach that’s apparently been effective in New York.
  • House – The far more aggressive House bill would require insurers to cover out-of-network emergency care at in-network rates, end the practice of balance billing (when the provider bills patients for medical fees that the insurer wouldn’t pay), and force insurers to make a minimum payment to out-of-network providers based on their in-network rates.

It’s generally safe to assume that most healthcare legislation won’t make it through Congress, but these new bills have a fighting chance, given their bipartisan sponsors, the general consensus that surprise medical billing is bad, and even presidential support.

AI for Ultrasound-Guided RT
A team of researchers from Stanford and Shandong Normal University demonstrated an ultrasound-guided radiotherapy system that uses a pair of machine learning algorithms to track tumor motion during radiotherapy. Although ultrasound-guided RT isn’t new, the system’s two complementary neural network architectures could make it a stronger alternative to image-guided X-ray or CT techniques:

  • The “first” algorithm extracts features from each image sequence frame and feeds this spatial information to the second algorithm
  • The “second” algorithm observes patterns over time to predict where the target will be in subsequent frames

When applied to a dataset of 64 liver ultrasound sequences (25 for training, 39 for testing) that were already evaluated by physicians, the algorithms achieved an average error rate of under 1mm and never exceeded the 2mm clinical threshold, while maintaining a 66 frame per second rate. The system also runs on an “unexceptional computer” with a dedicated graphics processing unit, reducing hardware-based barriers to clinical adoption.

The Wire

  • A survey of 200 radiologists and 70 radiology residents in France revealed a widespread lack of experience-with and understanding-of artificial intelligence, but high interest in learning more about AI. Only 7% of the surveyed radiologists use AI on a day-to-day basis and 73.3% report they have not received sufficient information about AI, but 94.4% would consider ongoing education on AI and 69.3% would attend “technically advanced” AI training. Meanwhile, 79.3% believe AI will have a positive impact on their future practice, which is good news in light of today’s lead story.
  • Despite the significant expansion of breast density notification laws, a study published in the Journal of Women’s Health (n = 155) found that many physicians are unaware of breast density laws (48%), feel they need needed more education about breast density and supplemental screening (67%), and are unaware that women with dense breasts have increased risk of breast cancer (62%). Compared to specialists, PCPs were less aware of breast density laws (32% vs. 71%), less knowledgeable about the increase in breast cancer risk with dense breasts (17% vs. 56%), and less likely to be “comfortable” answering patients’ breast density questions (8% vs. 55%).
  • MaxQ AI announced its new FDA/CE-cleared ACCIPIO Ax solution, which supports ICH triage through workflow prioritization and slice-level preview, and is highlighted by its ACCIPIO Ax SliceMap feature (guides users to CT slices with suspected ICH within a PACS viewer). ACCIPIO Ax will launch in August as the second component of MaxQ AI’s ACCIPIO ICH platform, joining its ACCIPIO Ix solution.
  • Research shared at ISMRM and originally reported by Auntminnie.com found that a GAN algorithm can be used to create synthetic contrast-enhanced brain MR images that “greatly resemble” images captured using gadolinium-based agents and can “predict some enhancing lesions that are not shown in the precontrast images.” The GAN algorithm was trained on images from 323 patients and tested with images from 81 patients (all a GE 3T MRI), producing synthetic images with a 0.939 similarity index (vs. 0.836 for precontrast images) and a 27.5 peak signal-to-noise ratio (vs. 19.6 for precontrast).
  • Here’s a far more traditional way to reduce GBCA exposure. A study out of the Hospital of the University of Pennsylvania detailed how an adjustment to its imaging protocols reduced its volume of contrast-enhanced scans for patients with multiple sclerosis. The hospital began performing real-time noncontrast FLAIR imaging assessments and only using contrast-enhanced MRIs if the initial assessment revealed evidence of new disease activity, allowing 87% of patients to avoid GBCA-enhanced imaging.
  • According to a new Medscape survey (n=20k physicians), 49% of radiologists have a net worth of $2 million or higher, equaling plastic surgery and orthopedics as the highest net worth specialties. It gets better, as 16% of radiologists have a net worth above $5 million (6th most), 16% radiologists have a net worth under $500,000 (tied for fewest), and 19% of radiologists are still paying off med school (4th fewest).

The Resource Wire

– This is sponsored content.

  • Carestream’s first OnSight 3D customer, Resurgens Orthopaedics (24 locations, 104 physicians), shared some of the benefits they’ve experienced from the cone beam CT system in this video, including the importance of weight bearing in surgery decisions, and the system’s image quality, ease of use, and fast study time.
  • 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.
  • The Focused Ultrasound Foundation’s 2018 Year in Review details the impressive research and clinical achievements that took place last year.

You might also like

You might also like..

Select All

You're signed up!

It's great to have you as a reader. Check your inbox for a welcome email.

-- The Imaging Wire team

You're all set!