“Patients note that radiologists can be held accountable for their mistakes, and they wonder who can be held responsible for errors made by computers.”
A team of Dutch researchers regarding patients’ perspective on the role of AI in radiology, and apparently their need for accountability.
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
Happy 100th Imaging Wire issue, everyone. Technically I wrote 23 of these while I was still building the site and newsletter, so the 100th issue to actually to hit inboxes won’t come for a few more months. Either way, this is a feel-good milestone. Big thanks to all of our readers who spend their mornings catching up on radiology news with us, thanks to the friends and family who’ve helped in so many ways, and a huge thanks to our newsletter sponsors. Please make sure to visit their sites and see what they’re up to. They make this publication possible. Onward to the next 100.
Agfa HealthCare’s Extended Independence
With the separation of its Health IT and imaging hardware business now complete, Agfa-Gevaert announced plans to “further extend the independence of Agfa HealthCare” and enlisted J.P. Morgan to advise this effort. In other words, the now-separated Agfa health IT division is likely headed towards some kind of acquisition or spin-off. The company assured that this move would give Agfa Healthcare “the strength and resources they need to pursue profitable growth,” which is pretty logical considering the challenges facing Agfa’s DR and print hardware operations, and given the opportunities Agfa HealthCare has now that it is exclusively focused on health IT. This move shouldn’t come as a major surprise to the industry’s Agfa watchers, as this was almost certainly Agfa’s goal when it announced plans to separate its health IT and radiology hardware business in 2017. That said, the next steps in Agfa HealthCare’s “independence” is harder to predict.
POCUS’ Heart Failure Detection Advantage
New research published in JAMA Network Open found that point-of-care lung ultrasound is as specific and more sensitive than chest X-ray in the identification of cardiogenic pulmonary edema, suggesting that lung ultrasound’s radiation, speed, and performance advantages might make it a superior method for evaluating dyspnea patients who are at risk of acute decompensated heart failure (ADHF). The researchers reviewed 6 prospective cohort studies (n=1,827 patients with ADHF), with their pooled estimates showing that lung ultrasound had a notable sensitivity advantage (0.88 vs. 0.73) and equivalent sensitivity (0.9) compared to chest X-ray. In other words, lung ultrasound could diagnose 15% more patients than chest X-ray without any increase in false positives. The researchers called for further studies into whether POCUS’ advantages would actually lead to improved diagnosis, treatment, and outcomes for patients with ADHF, but these initial results are pretty compelling.
Every once in a while, there’s actually a pro-EHR story… Research from a team of Duke University Medical Center radiologists found that transitioning from a PACS-driven workflow to an EHR-driven workflow “improved efficiency, was favorable to radiologists, and enhanced examination prioritization.” The researchers studied 16 radiologists’ CT exam order processes during the 30 days before and after the EHR workflow shift and performed a 12-question survey before (w/ 73 radiologists) and after (w/ 63 radiologists) the transition. The post-EHR improvements were pretty clear, including major reductions in time to open (17.3 ± 9.5 seconds vs. 52.4 ± 16.9), mouse actions (4.8 ± 1.5 vs. 9.5 ± 3.9), and keystrokes (0.1 ± 0.3 vs. 6.3 ± 2.9). Given those efficiencies, it’s not surprising that 54.7% of the radiologists rated the new EHR-driven workflow’s ease of use as “good” or “very good” (vs. 4.2% with the previous system), specifically highlighting satisfaction with accessing information for an exam (70.3% vs. 30%) and searching for relevant clinical info (81.3% vs. 19.2%).
AI Lung Cancer Predictor
In search of a solution to low-dose CT’s 96% false positive rate for lung cancer screening, a team from the University of Pittsburgh and UPMC Hillman Cancer Center developed a machine learning model that can effectively differentiate between benign and cancerous nodules. The researchers used the clinical, demographic, and LDCT data from 92 subjects to identify the three variables linked to malignant and benign cancer (number of nodules, number of blood vessels surrounding nodules, and years since the patient quit smoking) to build the Lung Cancer Causal Model (LCCM), which they then validated with data from 126 additional subjects. The study found that LCCM was more accurate than previous models and was able to identify 30% of the benign nodules that would have traditionally required further testing (including the costs, time, and discomfort that comes with these tests). The researchers seem bullish about LCCM, suggesting that integrating blood tests may add additional insights and revealing plans to test the model on a larger 6,000 scan dataset from the National Lung Screening Trial.
Patients Just as Ambivalent About AI as Radiologists, Maybe More
A new study published in JACR revealed a number of understandable insights into patients’ perspective on using AI in radiology. The study of 20 people scheduled for a CT scan (63.8 avg. age) found that patients generally prefer that their scans are read by human radiologists, although they could learn to trust computer diagnoses with sufficient evidence, are particularly open to AI-based second opinions, and believe AI could lead to time and cost savings. However, patients don’t fully understand how AI is applied to radiology (obviously), are concerned that AI will further reduce interactions with clinical staff, and are uncomfortable with AI being responsible for their healthcare or accountable if something goes wrong.
Share Your Feedback
We need your feedback to help make your Imaging Wire reading experience as valuable as possible. Please take three minutes to fill-out this 10 question survey to tell us what you like and help us find ways to improve.
- The Medicare Payment Advisory Commission’s (MedPAC) latest report to congress found that medical imaging procedure volume per Medicare beneficiary grew by 1.3% in 2017 (in units) after declining by 0.2% between 2012 and 2016. The 1.3% imaging increase was just below the average growth of overall medical services in 2017 (1.6%), but still contributed to a massive 75% increase in imaging procedures since 2000 that significantly outpaced growth of “major procedures” and “evaluation and management services” (47% and 45% since 2000). Imaging’s 2017 growth was driven by CT (+4.9% vs. +2.7% in 2012-2016) and MRI (+2.3% vs. +1.3% in 2012-2016), while X-ray (-0.1%), ultrasound (-0.1%), and nuclear imaging (+1%) were flat or declined in 2017.
- Fujifilm Medical Systems U.S.A. announced the FDA clearance and commercial launch of a trio of new image processing and software additions to its ASPIRE Cristalle DBT mammography system. The new features include: S-View (creates a synthesized 2D image from DBT, lowers dosage by ~50%), Iterative Super Resolution (ISR – a reconstruction process that lowers dosage and improves image quality), and Tomosynthesis Spot (focuses compression at a specific area); all of which first debuted at RSNA 2018.
- New research from Germany revealed that a pair of 68Ga-FAPI radiotracers may outperform 18F-FDG radiotracers for PET/CT cancer detection by directly targeting cancer-associated fibroblasts (CAFs – proteins that promote tumor growth, migration, and progression). In the study (n=50), the FAPI (fibroblast activation protein inhibitor) tracers matched or surpassed FDG PET/CT’s image quality, with substantially shorter image acquisition times (10min vs. 1hr) and without requiring patients to restrict their diet before imaging.
- Siemens Healthineers marked the annual ACC conference with the launch of the SOMATOM go.Top Cardiovascular Edition, adding personalized dose control and advanced cardiovascular capabilities to the 128-slice CT system. Targeting both hospitals and outpatient cardiology offices, the new SOMATOM go.Top Cardiovascular Edition supports a range of cardiovascular features (e.g. CCTA), but is highlighted for its inclusion of HeartFlow FFRCT Analysis, a personalized cardiac test from super-funded AI startup Heartflow, that helps clinicians evaluate the impact a blockage has on blood flow and determine the best treatment for patients.
- A new Susan G. Komen-funded study found that women within suspicious findings from screening mammograms are delaying or even declining follow-up diagnostic exams because of high out-of-pocket costs. The study (n=30) found that high and varying costs lead to “frustration, confusion and sometimes avoidance” of diagnostic exams and the average out-of-pocket costs for a diagnostic mammogram ($234) or a breast MRI ($1,021) are particularly burdensome for people who have already been diagnosed with breast cancer, since diagnostic tests are recommended for them instead of traditional screening.
- Sonavex announced the FDA 510(k) clearance of its EchoSure system, which combines 3D ultrasound imaging with deep learning algorithms to automate post-surgery blood flow monitoring. When used with Sonavex’s FDA-cleared EchoMark bioresorbable markers (ultrasound-reflective implants), EchoSure can reportedly measure blood flow without ultrasound expertise, while surgeons can monitor patients remotely from their mobile devices using its EchoSure App.
- A team of French researchers repurposed an algorithm that was originally developed for a 2009 Netflix movie preference prediction competition to speed up the acquisition of Raman spectroscopy images from several minutes to about 30 seconds. The new “Compressive Imaging” computational imaging approach acquires less Raman spectral data than normal and then uses the matrix completion algorithm to fill in the remaining information to the point that the image could be used for diagnosis.
- Australian enterprise imaging company, Mach7 Technologies, announced a major restructuring and cost cutting effort intended to help the company break even within the next 12 months, while also driving revenue growth and improving service delivery. The cost cutting effort largely targeted Mach7’s executive ranks, bringing the immediate departure of CEO Mike Jackman (temporarily replaced by COO Mike Lampron) and the transition of CTO Eric Rice to a consulting role, while shifting some functions to lower-cost regions or outsourced providers. Meanwhile, founder and CSO Ravi Krishnan will take on an expanded role in U.S. sales, in addition to his previous focus on Asia-Pacific and the Middle East, and the company will look to add sales and service roles in the US and Asia.
- A team of Spanish researchers found that medical students and family physicians view using the Second Life online training platform to be a valuable and enjoyable way to improve their understanding of radiology. The study’s 48 third year medical students rated their Second Life training with a 4.2 average score (out of five), while the 14 family physicians gave it an unspecified lower average score and had lower participation rates, which the researchers suggested may be related to the fact that the students had six of their classes at the same time and the PCPs’ training was completely independent. This study is the latest to highlight the benefits of online training, coming just a few weeks after an Australian study revealed even more impressive clinical results.
- French breast imaging AI company, Therapixel, completed a €5 million Series A round that it will use to add to its staffing and fund the development and commercialization of its breast cancer screening solution. Therapixel hasn’t made much noise over the last year, but it did win a 1,200-participant digital mammography competition in 2017, and claims to be able to reduce false-positives by 5%.
The Resource Wire
–This is sponsored content.
- Bellin Health Titletown Sports Medicine & Orthopedics of Green Bay, Wisconsin (located right next to Lambeau Field) became the latest orthopedic practice to install Carestream’s OnSight 3D Extremity System, specifically highlighting the cone beam CT’s image quality and performance during weight-bearing exams.
- This Focus Ultrasound Foundation blog post from ultrasound pioneer, Narendra Sanghvi, looks back on the modest beginnings and early development of focused ultrasound.
- POCUS Systems is approved as a Veteran Owned Business with the US Government Office of Veterans Business Development, paving the way for partnerships with the federal healthcare delivery systems.
- This Medmo video details how its healthcare marketplace platform and network of participating radiologists help underinsured patients pay as little as possible for their imaging procedures.