“If AI earns a place in breast imaging, its best opportunity is to decrease false-positive flags.”
A pair of MD Anderson Cancer Center radiologists.
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
Agfa DR 800 Adds Another Function
The capabilities of Agfa’s “multi-purpose” DR 800 became even more diverse with the FDA clearance of the DR 800 With Tomosynthesis, adding the image reconstruction technique to the system’s existing support for radiography, fluoroscopy, and advanced clinical applications. The system’s MUSICA Digital Tomosynthesis software, which was first introduced before RSNA, features tomosynthesis algorithms that produce multi-slice images with less noise and fewer artifacts in under a minute. Even before adding tomosynthesis, Agfa has placed an extremely high level of emphasis on the DR 800’s multifunctional capabilities, using swiss army knife imagery across its product materials and emphasizing versatility as the core of the system’s value proposition.
The Keys to AI Success
An editorial from pair of MD Anderson Cancer Center radiologists outlined the four areas where AI must excel in order to “earn a role” in women’s imaging: performance, time, workflow, and cost. The authors highlighted mammography performance as the greatest AI-related requirement, specifically emphasizing the need to reduce false positives compared to the current performance of CAD programs. Efficiency was next on their list, demanding that AI must not increase interpretation times or delay the availability of images to radiologists. The third item was a bit more futuristic, calling for AI to seamlessly integrate into clinical workflows, including leveraging EHR records to ID patients that require imaging, automatically ordering and scheduling the exam on behalf of physicians, and then supporting radiologists’ diagnoses. Finally, the radiologists emphasized that the cost of AI must not be so high that it outweighs the value it provides. This is a solid list, and not just for breast imaging.
Screening’s Better with DBT
A study from team of Norwegian researchers (n=24,000) found that the inclusion of digital breast tomosynthesis (DBT) in breast cancer screening programs resulted in significant sensitivity and specificity improvements. The study also found that synthetic mammography (SM) may be a viable alternative to digital mammography (DM) when combined with DBT, while finding that adding CAD to DM did little to improve screening effectiveness, especially in comparison to DM+DBT.
- DM Alone: 54.1% sensitivity, 94.2% specificity
- DM+CAD: 56.2% sensitivity, 94.2 specificity
- SM+DBT: 69% sensitivity, 95.4% specificity
- DM+DBT: 70.5% sensitivity, 95% specificity
This study certainly isn’t the first to highlight the benefits of DBT+DM compared to just DM (here’s another example of DM+DBT’s recall rate advantage) or SM+DBT’s potential as a DM+DBT alternative (here’s another example of that), but it certainly helps support these two arguments.
MIT’s Neuroimaging Breakthrough
MIT researchers developed a new MRI sensor that can image calcium activity deep within the brain, potentially leading to new understanding of how neurons communicate and new insights into neurons’ roles in human behaviors. The MIT team has been working on MRI-based calcium imaging for some time, achieving the ability to measure calcium on the outside of brain cells last year and is now able to image calcium within brain cells due to a new manganese-based contrast agent. This may be word soup to Imaging Wire readers who don’t happen to be neuroscientists. However, this appears to be a significant neuroimaging breakthrough, imaging far deeper into the brain than current microscopy-based calcium imaging techniques and potentially providing far more precise data on the location and timing of neuron activity compared to traditional functional MRI (fMRI) imaging.
Breast MRI Referrals Helped by Legislation and Education
Research from Beth Israel Deaconess Medical Center and Harvard found that although referrals for breast MRI remain rare, a combination of breast density legislation and breast MRI education can drive adoption. The study (n=134) revealed that a respectable 42% of referring physicians actively order breast MRIs, 30% of those physicians have increased their breast MRI referrals since Massachusetts began requiring density notifications in 2015, and 29% would make even more breast MRI referrals for women with dense breasts if costs were lower. However, the overall volume of breast MRI orders is still low, as 79% of MRI-comfortable referrers order fewer than 10 exams per year. Among the physicians who don’t order breast MRIs, 42% cited a lack of familiarity with the modality, prompting the researchers to encourage efforts to educate referring physicians on the utility of breast MRI.
- ViewRay announced the FDA clearance of new soft tissue visualization features for its MRIdian MRI-guided radiation therapy system. The enhancements to the system’s SmartVISION MRI imaging technology include improved visualization and contrast between different tissues (supports tissue visualization and beam contouring), a new diffusion-weighted imaging feature (supports assessment and prediction of tumor response to RT), improved anatomical planar imaging performance (allows 2x increases to imaging speed, resolution, and MR signal-to-noise ratio), and faster MLC speeds (potentially reducing treatment times).
- Research from the VCU health system revealed that 70% of women (n= 2,245) prefer to receive their mammogram screening results within 48 hours of their exam and 55% percent prefer to be contacted about their results via a phone call (followed by text, letter, and then email), suggesting that prompt and personal notifications are key to improving patient-centricity in this part of the screening process.
- Kaiser Permanente announced that it is waiving all tuition and fees for its new medical school’s first five graduating classes (students entering 2020-2024), saving each student an estimated $275,000 over the course of the four-year program. Like NYU’s recent tuition-free move, Kaiser Permanente was motivated to reduce the role of debt on potential doctors’ decisions on whether to study medicine as well as the impact of debt on med students’ career and location choices.
- Dutch medical imaging AI developer, Aidence, generated €10 million from a series A funding round, bringing its total funding to €12.5 million ($11.3m, $14.2m). Aidence will use the new funding to expand its Veye Chest lung cancer detection solution across Europe, support the solution’s FDA approval process, and fund further development.
- A team of University of Washington researchers tackled lung cancer screening’s participation problem in a recent JACR paper, which is apparently significant given that only about 5% of people who are eligible for LCS participate in such programs. The team started by defining LCS’ problems, including inadvertently screening ineligible participants, an insufficient effort to encourage smoking cessation, and omitting patients of lower socioeconomic status. The team’s solutions had a similar focus, including investing in patient navigation roles/practices (for participant outreach, eligibility evaluation, and decision support) and ensuring that participants are eligible based on their actual smoking history (and not necessarily what their EHR records say about smoking).
- MaxQ AI announced that its collaboration with Intel helped the company triple the computing power of its Accipio lx intracranial hemorrhage (ICH) and stroke detection platform, without impacting detection accuracy. Many of the Accipio platform’s algorithms are written using the Intel Distribution of OpenVINO Toolkit and are optimized for Intel Processors.
- Carestream announced that Canada’s University Health Network purchased 10 DRX-Evolution Plus DR systems and eight DRX-Revolution Mobile X-ray Systems, expanding an initial deal for a single DRX-Evolution system a year ago.
- UK Research and Innovation (UKRI) and a number of public and private sector partners (AstraZeneca, Google and Rolls-Royce, and NHS trusts) are investing £200 million ($261m) to support the education of more than 1,000 of the country’s future AI leaders at 16 new Centres for Doctoral Training (CDTs) located at 14 UK universities. The UK government and the NHS have prominently touted ongoing and future investments in AI during the last year, often with a focus on healthcare, but this appears to be the most significant AI investment in the country to date.
- PaxeraHealth will launch its new PaxeraShare universal image sharing platform at this week’s ECR 2019 conference, which can be deployed into any existing PACS system “in a few hours,” and supports secure provider-to-provider and provider-to-patient image sharing.
- University of Alberta researchers developed an AI algorithm that can accurately identify lung tumors in MRI scans and produce patient-specific models to support surgical planning. The researchers trained the neural network on a set of 600 MRI scans featuring labeled lung tumors, outperforming “other advanced techniques” in tests.
The Resource Wire
–This is sponsored content.
- The Focused Ultrasound Foundation published its 2018 Year in Review, detailing the organization’s impressive research and clinical achievements over the last year.
- POCUS Systems’ forthcoming ultrasounds will combine ease of use, durability, and reliability, allowing clinicians to focus on their patients.
- How much does a CT scan cost? According to Medmo, that depends. Scans made with the exact same device on the exact same body part could cost $225 at one facility and $2,500 at another. Medmo also provides some advice to make sure patients don’t pay too much for their scans, including using the Medmo Marketplace, where the average CT costs between $225 and $700.
- Carestream highlights the ongoing adoption of its OnSight 3D Extremity Systems, which provide orthopaedic specialists with high-resolution 3D images for more accurate diagnostic and treatment decisions, providing testimonials from a number of OnSight 3D customers.