The Killer App | More than Notifications | Three Steps to Independence

“To address these unquestioned problems and make healthcare great again, one should not lock his thoughts within the walls of the radiology department and rather seek relentlessly for the most meaningful problem to solve: the Killer App!”

Amine Korchi, MD encouraging AI developers to look beyond the pixel if they want to MHCGA.


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


The Killer App

A new Medium post from Amine Korchi, MD spared few in radiology AI, questioning the value of most AI solutions, the viability of many AI business models, radiology AI’s ability to scale, and even whether today’s solutions actually qualify as products. If you’re among the AI sceptics out there or are fully invested in AI but appreciate some tough love, then this one’s for you.

  • Narrow AI – Dr. Korchi echoed others in criticizing the industry’s current focus on narrow AI algorithms that detect one to several abnormalities on a single imaging modality and are based on limited training datasets.
  • The Data Problem – Korchi reminded everyone that most AI startups are building algorithms with data from hospitals in their geographical region (therefore small and non-representative) and rely on a human-driven annotation process that is inherently unscalable.
  • Feature vs. Product – As a result, Korchi suggests that today’s AI players are largely offering nice-to-have features (that support a task) rather than need-to-have products (that complete a job), warning that AI companies will have to produce complete products to be successful.
  • Pixel Focused – Even though image analysis is the reason radiology is at the center of healthcare AI, Dr. Korchi claims that the current focus on “image pixels” may be addressing healthcare’s symptoms (e.g. prioritizing worklists for burnt-out rads) rather than its problems (e.g. proactively identifying the people who actually need imaging).
  • The Killer App – With that, Korchi called for the industry to look beyond the walls of the radiology department and seek most meaningful problem to solve, or as he calls it “the Killer App.” He actually discussed solutions in this part of the post, using Intuitive Surgical’s Da Vinci system as an example of a killer app that succeeded as a result of its clear clinical benefits.
  • Algorithm Commoditization – Dr. Korchi forecasted quite an AI storyline plot twist, suggesting that the “algorithm is on its way to become a commodity” (not radiologists as some have suggested), noting that the tools to build AI are open source and AI skills can be learned. With data science eventually becoming commoditized, he argues that academia’s massive data access advantages would allow them to produce better algorithms than commercial AI firms.

This is among the most scorching takes on radiology AI we’ve seen from someone with the credentials and reach of Dr. Korchi, so even if many of these points have been said before, this quickly-circulating post surely has some people looking at today’s radiology AI market in a new light.



It Takes More than Notifications

Research from the University of North Carolina reveals that breast density notification laws might not lead to increased supplemental screening without the right healthcare environment to support it. Here’s what they studied and found out:

  • The team studied screening data before (2012-2013) and after (2014-2016) North Carolina adopted its density notification law.
  • Supplemental screening use was similar before and after the notification law, regardless of breast density.
  • Although screening volumes didn’t increase, DBT screening increased by 600% among women with dense breasts, growing much faster than DBT screening among women with non-dense breasts.
  • Noting that other states experienced significant increases in supplemental screening after passing breast density notification laws (CA, NJ, TX), the researchers suggested that North Carolina’s healthcare coverage laws (does not require supplemental screening coverage) and limited MRI availability may have hindered adoption.

The UNC team suggested that other states and policymakers should consider the Tar Heel state’s notification law results to making sure their environment supports the goals of their screening policies.



Three Steps to Independence

A recent Healthcare Administrative Partners (HAP) blog post outlined a path for radiology practices to stay independent, noting the growing consolidation pressures within healthcare and many radiologists’ desire to resist this trend. HAP isolated three core reasons that practices choose to merge: economics (e.g. capital for upgrades, negotiation leverage, senior partner exit plans), quality improvement (e.g. the need to meet quality improvement plans, subspecialist access), and practice governance (e.g. partners lacking management skills, insufficient access to IT experts), however there are other options:

  • Build Relationships – Especially with the hospitals that a practice serves, including participating in hospitals’ physician hospital organization (PHO) to access lower contract rates, working with hospitals in payor negotiations, and bundling radiology with other services that hospitals can negotiate.
  • Get Help – As might be expected, HAP encourages “judicious use of outside support,” such as leveraging teleradiology services, contracting experienced practice managers, and working with management service companies to support operations.
  • Get Capital – “Capital for upgrades, expansion, or infrastructure is generally available through financing” and “makes a lot of sense” for specific purposes such as acquiring a revenue generating imaging system or building a solid IT infrastructure.

As HAP puts it, these investments are valuable for any practice trying to stay independent, and even if they don’t end up keeping their practice independent, these steps should make them a more marketable acquisition target.



Proton Imaging Breakthrough

A Chicago area research team is developing a proton CT and radiography imaging system that they believe will help make proton therapy guidance and planning far more precise than the current X-ray-based CT method. Here’s how it would work:

  • Direct Stopping Power – Unlike current CT-based planning that “introduces inherent inaccuracies due to the need to convert CT Hounsfield units into proton relative stopping power,” the Proton CT system measures proton stopping power directly, significantly reducing beam range uncertainties.
  • pDRR – The Proton CT system could also generate proton digitally reconstructed radiographs (pDRRs) that can potentially replace X-ray-based DRRs in the daily image-guidance process, with just 1% of the X-ray dosage and greater accuracy.
  • Early Results – The team shared the first proton radiographs of a biological specimen (a frozen tilapia), achieving detailed images that would be useable for daily image-guidance, and revealed that it has also obtained its first proton CT images. They also developed an image reconstruction system for the proton radiography system that was able to display most features found in a ground truth image, with plans to develop a proton CT reconstruction system.
  • Next Up – Although they face many hardware and software challenges, the team is now designing a clinical version of the proton imaging system and entered a partnership with Cosylab to integrate its new imaging system into proton therapy treatment rooms.

The Wire

  • ASRT’s 2019 radiologic technologist wage survey (n=14,165) revealed that the average RT salary increased by 5.3% over the last three years to $69,266, which still lagged behind the same period’s 6.8% inflation rate. Wages increased across all RT disciplines, with the greatest growth among nuclear medicine technologists (+10% to $83,385), radiation therapists (+7.7% to $89,159), and MRI technologists (+7.2% to $76,177), while medical dosimetry ($114,891) and radiologic assistants ($108,494) had the highest average salaries.
  • Researchers at South Korea’s Center for Quantum Nanoscience developed an MRI system that is able to visualize the magnetic field of single atoms. The team modified the tip of a Scanning Tunneling Microscope so it spun like an MRI system, allowing it to map an atom’s 3D magnetic field with high resolution and distinguish different kinds of atoms by their magnetic field signature. Although this breakthrough may have some medical contributions (e.g. drug discovery), the researchers seemed more excited about what it means for quantum computing.
  • A new report from healthcare recruiter Merritt Hawkins (n = 3,131 physicians and advanced practitioners) revealed that 78% of its 2018 recruiting assignments were for specialists (vs. 67% in 2015), while recruiting assignments for PCPs fell by 38% over the same period. Radiology was the fourth most requested practice area for the second straight year (after family medicine, psychiatry, OB/GYN), helping to drive radiologists’ average base salary up 4.3% to $387,000 ($309k – $650k range).
  • A team from Duke and Washington University developed an AI algorithm that could support radiologists’ thyroid nodule biopsy recommendation decision making. The deep learning algorithm was trained on data from 1,377 thyroid nodules from 1,230 patients (ultrasound images and cytologic or histologic diagnoses) and then tested with 99 thyroid ultrasound images against a panel of ACR TI-RADS radiologists, achieving 87% sensitivity (13 of 15, matching expert consensus, and beating 5 of 9 rads) and 52% specificity (44 of 84, similar to expert consensus, and beating 7 of 9 rads).

The Resource Wire

– This is sponsored content.

  • Carestream’s AHRA 2019 booth will spotlight its ImageView software and mobile imaging portfolio, including its CARESTREAM DRX-Revolution Nano Mobile X-ray System, CARESTREAM DRX-Transportable System/Lite, and new CARESTREAM DRX Plus 2530C Detector.
  • By partnering with Medmo, imaging centers can keep their schedules full and their equipment busy. Here’s where to learn more and get started.
  • This POCUS Systems video details how its forthcoming ultrasound helps patients and providers, and reduces reliance on MSK MRI use.

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.