Artificial intelligence equivalency in breast cancer screening; diffusion-weighted imaging and breast MRI; CT patterns for EVALI; Emergency Department and CT Scan Declines with COVID-19; and Thoracic Imaging Findings of MIS-C.
Welcome to Diagnostic Imaging’s Weekly Scan. I’m Whitney Palmer, senior editor.
Before we get to our featured interview this week with Dr. Einat Blumfield, attending radiologist at Children’s Hospital at Montefiore about the thoracic imaging findings of multi-system inflammatory syndrome in children or MIS-C and what radiologists need to know, here are the top stories of the week.
It looks like it might finally have happened. Artificial intelligence has reached an equivalency point with radiologists. To be clear, it’s one artificial intelligence algorithm tested in one study focused on breast cancer screening. But, the results are promising, according to investigators from Karolinska Insitutet. They tested three AI algorithms that were designed to identify breast cancers based on previously captured mammograms to see which one performed the best. They published their results in JAMA: Oncology. To make this determination, they examined mammograms from 8,805 women between the ages of 40 and 74, 739 of whom have a breast cancer diagnosis. Based on their analysis, the team determined an algorithm that was trained with 72,000 cancer images and 680,000 normal images performed with the same accuracy as the average radiologist. It had 81.9-percent sensitivity, 96.6-percent specificity, and an area under the curve of 0.956 for the detection of cancers either at screening or within 12 months. These results, they said, were favorable in comparison to the U.S. Breast Cancer Surveillance Consortium benchmarks of 86.9-percent sensitivity and 88.9-percent specificity.
Work is also continuing to improve the use of diffusion-weighted imaging with breast MRI. Widespread use has been slower to take off due to concerns over spatial resolution and image quality, and better spatial resolution is needed to identify the changes in cell density, organization, and membrane integrity that accompany small breast lesions. But, a new proof-of-concept study published in Radiology from researchers at the University of Minnesota has shown that the axially reformatted simultaneous multi-slice protocol (AR-SMS) can produce higher image quality than either standard SE echo-planar imaging or RS echo-planar imaging. By testing these three 5-minute protocols on a 3T scanner on 30 lesions from 28 women, the team determined AR-SMS outperformed both RS echo-planar and SE echo-planar imaging in both image quality and rank. On the Likert scale for image quality, AR-SMS scored 0.74 points higher than RS echo-planar and 1.31 points higher than SE echo-planar imaging. It also achieves larger anatomic coverage and better image quality. Still, the researchers said, it requires robust distortion correction and is not yet commercially available. Consequently, they suggested combining the encoding speed from AR-SMS and the reduced distortion of RS echo-planar imaging for future work.
COVID-19 has rightfully captured radiology’s attention this year, but it is not the only respiratory disease that has radiology's focus. New evidence was published this week in Radiology: Cardiothoracic Imaging that reveals lung injuries caused by vaping or e-cigarette use, called electronic cigarette or vaping product use-associated lung injury or EVALI, leave signature marks on CT scans. Because the symptoms of this disease are relatively non-specific, including shortness-of-breath, chest pain, and other symptoms that could indicate a viral infection, these findings could help radiologists make faster, more accurate diagnoses in younger people, as well as help them avoid unnecessary biopsies. A team from the Mayo Clinic conducted this study with 26 patients who satisfied the criteria for EVALI – they had vaped within 90 days of their symptoms appearing, they had chest imaging abnormalities, and any other diagnoses had been eliminated. Upon examining their CT scans, the team found a common injury pattern: ground-glass opacity and consolidation. The second most common pattern resembled subacute hypersensitivity pneumonitis. An acute eosinophilic pneumonia-like pattern and organizing pneumonia-like pattern were much less common. The hope, the team said, is that radiologists will be better able to identify EVALI and pass their concerns on to the referring physician who can then conduct a nicotine metabolite test or ask the patient further questions about exposure.
In COVID-19 news this week, researchers from the Universities of Ottawa and Toronto published their findings that fewer people are heading to the emergency room during the pandemic, leading to a decline in abdominal CT scan and worsening prognoses for patients. In the Journal of the American College of Radiology, the team reported a 46.7-percent drop in the net number of emergency room visits and a 42-percent decline in abdominal CT scans between the same 4-week period in 2019 and 2020. For the same time period, the number of patients with complications and those who required surgery rose. To make this determination, they examined abdominal CT scans from patients who were referred to radiology from the emergency room. The number fell from 733 in 2019 to 422 in 2020, but the number of scans with positive findings rose from 32.7 percent in 2019 to 50.5 in 2020. In addition, complications increased from 7.9 percent to 19.7 percent, and patients needing surgery increased from 26.3 percent to 47.6 percent. The reasons behind this shift aren’t exactly clear, but the team postulated that people were putting off coming into the emergency room because they want to avoid potential exposure to the virus, and when they finally do come in, their conditions are far worse, leading to a poorer prognosis.
And, finally this week, Diagnostic Imaging spoke with Dr. Einat Blumfield, an attending radiologist at the Children’s Hospital at Montefiore about her recent investigations into the thoracic imaging findings of multisystem inflammatory syndrome-in children or MIS-C. Not only did she share with us the results of her investigations, but given the time of year with schools beginning to re-open, she said these findings support slow, methodical school re-openings that come with plans to scale back in COVID-19, once again, begins to surge. Here’s what she had to say.