AI’s Translational Roadmap | An AI Platform Play | Massive Reform

“I don’t have feelings and I can’t read, but I do know what you and your colleagues have been writing about me.”

An excerpt from EHR’s “personal letter” to clinicians (authored by a team of Denver physicians), intended to acknowledge their frustrations, share some tips on how they can coexist, and emphasize the benefits of working together. This is probably the safest way to stick up for EHRs these days.

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’s Translational Roadmap

A month after a group of radiology and AI leaders published medical imaging AI’s foundational research roadmap (mainly focused on technologies and processes), the same NIH-led team released a companion roadmap intended to accelerate translational AI research. The way they see it, translating foundational AI research to routine clinical practice will require a focus on these four priorities:

  • Creating structured AI use cases by defining and highlighting clinical challenges to solve
  • Establishing methods to encourage data sharing for AI training/testing to promote generalizability and mitigate bias
  • Developing tools for AI validation and performance monitoring to facilitate regulatory approval
  • Creating standards and common data elements to seamlessly integrate AI tools into existing clinical workflows

These priories make a lot of sense and we’ve heard flavors of each before, so it’s no surprise that this latest roadmap didn’t face many disagreements. Still, agreement doesn’t guarantee cooperation and getting AI’s diverse groups of developers, researchers, and users to organize under this plan may require its own roadmap.

Major Reform

The U.S.’ major healthcare cost themes (surprise billing, cost transparency, and drug costs) converged last week when the Senate HELP Committee released details on a massive piece of proposed legislation that may affect the entire healthcare industry. As one healthcare lobbyist put it, “folks should take this package seriously,” and here’s why:

  • It’s a bipartisan package produced by a chairman and a ranking member
  • It targets surprise billing by requiring independent arbitration for disputed bills, an in-network billing guarantee for any clinicians working at an in-network hospital (like radiologists), and that labs and diagnostic tests (like imaging scans) are billed at in-network rates.
  • It tackles prescription drug costs, making it easier to launch generic drugs and ensuring that pharmacy benefit managers pass savings on to customers

This package was a long time in the making but its next steps may come quickly, as the committee plans to advance it to the Senate floor by July. From there, the bill will face its share of challenges, but considering the support behind most of these reform targets, it’s very possible that the bill will at least serve as a key step in the U.S. government’s path towards healthcare cost reform.

Qure.ai and Incepto’s Platform Play

Qure.ai and French healthcare AI company, Incepto, publicly announced an already-active distribution agreement that made Qure.ai’s qER head CT and qXR chest X-ray solutions available in France, Belgium, and Switzerland for the first time. Although Qure.ai and Incepto have been working together for a while, this is still a significant announcement for both companies, and perhaps the industry.

  • For Qure.ai: The alliance bolsters its European presence, noting that Qure.ai’s direct teams are in talks with providers across the continent (Sweden, Germany, Switzerland, Italy, and more), but Incepto is its first European distributor.
  • For Incepto: The deal helps advance its goal of developing a “Netflix of AI” platform that’s intended to connect AI companies and hospitals. That’s a lofty goal, but it’s starting to take shape, as Incepto now has AI distribution agreements with ScreenPoint (mammography) and Qure.ai (head CT and chest X-ray), along with a trio of in-development solutions co-created with local hospitals (aortic aneurysms, intestinal obstruction, knee MRI).

There are of course other AI marketplaces and platforms available, with more on the way, but this is an interesting play for both Qure.ai and Incepto, and it’s worth keeping an eye on as the delivery/distribution part of the AI ecosystem takes shape.

The Wire

  • A Duke-led team developed an AI algorithm to improve the ACR TI-RADS risk stratification of thyroid nodules, making improvements to ease of use and specificity (two of TI-RADS’ main challenges). The team trained the “AI TI-RADS” algorithm with scored images from 1,325 biopsy-proven thyroid nodules and tested it on 100 nodules, achieving higher AUC (0.93 vs. 0.91) and specificity (65% vs. 47%) using data from expert radiologists, while also achieving higher specificity (55% vs. 48%) using data from non-expert radiologists. Although this may not get as many headlines as other AI studies, validating what findings are and aren’t useful seems like important work.
  • EMvision Medical Devices signed a co-development deal with Keysight Technologies Malaysia to produce a Vector Network Analyzer (VNA) component that will be used in EMvision’s next generation portable brain scanner. EMvision will leverage the new VNA to reduce the brain scanner’s size, which is crucial for first responders and point-of-care providers who need to quickly assess stroke and brain injuries.
  • A new paper published in Military Medicine detailed the successful results of implementing a POCUS ultrasound curriculum at a large military internal medicine residency program. The POCUS training began with a first-year pilot (voluntary, five 60-minute monthly courses) and then incorporated POCUS into its core curriculum during the second year (seven 3-hour monthly courses). The study found that trainees achieved modest improvements during the informal pilot program, but the structured second year program (n=75) drove significant improvements in ultrasound-guided procedure confidence (67.8% vs. 82.1%) and the proportion of respondents who anticipate using ultrasound in their clinical practice (63.6% vs. 81.8%).
  • A team of interstitial lung disease (ILD) experts announced their formation of the Open Source Imaging Consortium (OSIC), a global not-for-profit cooperative effort focused on improving ILD research. OSIC members will work together to create 15,000 anonymized images (1,500 by the end of 2019), develop machine learning algorithms based on the scans, and then work to incorporate the algorithms into commercial analysis tools used to perform imaging-based ILD diagnosis, prognosis, and prediction of therapy response.
  • Signify Research highlighted predictive analytics as “the next evolution” in medical imaging AI, suggesting that AI tools will deliver the greatest value once they have the ability to detect (the first evolution stage), quantify (the current stage), and predict (the next stage) conditions and outcomes. Once medical imaging AI solutions add predictive analysis, Signify suggests that they will allow even earlier detection and diagnosis, allowing radiologists to prioritize urgent cases and determine whether invasive follow-up procedures (like biopsy) are necessary.
  • A NYU Langone Health team used the SimpleNLP open source NLP tool to identify incidental lung nodules (ILNs) in radiology reports for assessment of management recommendations. The tool was trained on 950 unstructured chest CT reports reviewed for ILNs, later identifying ILNs with 91.1% sensitivity and 82.2% specificity in a validation set, with 59.7% and 97% respective positive and negative predictive values. Following the introduction of a Fleischner reporting macro there was no statistical difference in the proportion of reports with ILNs (21.6% vs. 22.4%) or reports with ILNs containing follow-up recommendations (69.4% vs. 79.2%).

The Resource Wire

– This is sponsored content.

  • In this Carestream video, orthopaedic surgeon Dr. Bryan Den Hartog presents clinical images illustrating traditional CT vs. extremity CT imaging and discusses how the image resolution in the OnSight 3D Extremity System helps in his practice.
  • A new study in JACR revealed that the rise of high-deductible health plans has led to greater patient concerns over imaging costs than ever before, while patient cost comparisons often leads to “confusion, misinformation, and opaqueness.” These are the exact patients who can be helped by the Medmo platform, which connects high-deductible patients with radiology centers, ensuring the best value for patients and a profitable revenue stream for imaging centers.

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!