This week in Radiology

This AudioScholar briefing is generated by artificial intelligence for healthcare professionals and trainees. It is not medical advice.

This week: Lung-RADS 2022 and the lung cancers that hide in cystic airspaces, gadolinium measured directly in children's bone, and estrogen receptor PET changing how breast cancer is managed.

Welcome to This Week in Radiology. We're covering 10 papers this week: 3 in depth, 4 in brief, 3 quick signals.

In Radiology, Mascalchi and colleagues pooled participants with lung cancer from three screening trials: the National Lung Screening Trial, ITALUNG in Italy and LUSI in Germany. This was a retrospective secondary analysis of 1258 participants. Two radiologists jointly reviewed the low-dose CT images to find cancers associated with cystic airspaces [1]. They found 73 such cancers, about 6 percent of screening-detected and just over 5 percent of non-screening-detected cancers. Multilocular lesions were more often advanced than unilocular ones: half of the multilocular cancers were advanced, against under a third of the unilocular cancers. Five-year lung cancer-specific survival was about 51 percent, close to the 53 percent seen in noncystic non-small cell cancer, with no difference after adjustment. When the Lung-RADS 2022 criteria for atypical cysts were applied, about three quarters of these cancers might have been diagnosed earlier. The main limitation is that this earlier diagnosis is hypothetical: the criteria were applied in hindsight, by readers who knew a cancer had developed, and the cystic group is small. The study adds a multinational estimate of how often this pattern occurs, and early support for the 2022 cyst criteria.

Also in Radiology, Roberts and colleagues ran a prospective pilot study at a single academic centre in children under 18 who were having a median sternotomy [2]. Bone removed at surgery was analysed for gadolinium in 29 children: nine exposed to the linear agent gadobenate, eleven exposed to the macrocyclic agent gadoterate, and nine never exposed. Gadolinium was found in every group, including the controls. Bone concentrations after gadobenate were roughly 130 times those in controls and about 16 times those after gadoterate. Gadoterate exposure was still linked to concentrations about eight times higher than in controls. The authors suggest the trace levels in controls may reflect environmental exposure. The limitations are substantial: the sample is small, the controls were much younger, with a median age under two years against early teens in the exposed groups, and all the children had cardiac surgery. No clinical consequence of retention was measured. The study adds the first direct measurement of retained gadolinium in paediatric bone, and it shows a large gap between the linear and macrocyclic agents studied.

In European Radiology, Vargas and colleagues report a systematic review and meta-analysis of fluoroestradiol PET/CT in estrogen receptor-positive breast cancer [3]. Ten studies with 773 patients were pooled. Management changed after the scan in 38 percent of patients compared with conventional workup. About one in six patients changed treatment within the same modality, and about one in ten switched modality. About 17 percent of patients moved from endocrine to non-endocrine therapy, and about 12 percent moved the other way. Among patients whose management did not change, the scan improved diagnostic certainty in close to four in ten. The key limitation is the endpoint: a change in management is a surrogate, and the abstract reports no survival, progression or quality-of-life outcomes. Several subgroup estimates were also imprecise. The analysis adds a pooled figure for the clinical impact of receptor imaging, which until now rested on scattered single studies.

Now, four in brief.

In the Journal of the American College of Radiology, Gareen and colleagues compared what happened after lung screening in 2018 Medicare beneficiaries with what happened in National Lung Screening Trial participants [4]. Within six months, Medicare beneficiaries were less likely to have follow-up imaging, at about 17 percent against 21 percent. They were more likely to undergo an invasive diagnostic procedure, at nearly 5 percent against under 1 percent. Surgical procedures followed the same pattern. At one year, the invasive and surgical differences were no longer statistically significant. Mortality did not differ, and the trial diagnosed more cancers. The comparison relies on imputed trial data and spans different eras, so it describes a gap between trial and community practice without explaining it.

In Investigative Radiology, Barr and colleagues built E-BI-RADS, a BI-RADS-style classification for breast elastography [5]. It combines shear-wave velocity with the ratio of elastography to B-mode lesion size. In a retrospective study of 344 lesions, sensitivity stayed near 100 percent, matching B-mode BI-RADS. Specificity rose from about a third to about three quarters. No malignancy occurred in the 71 lesions placed in category 2. The single missed cancer was a 31 millimetre lymphoma. The evidence is early: the study used one vendor and one transducer, included a high proportion of malignant lesions, and the abstract does not describe external validation.

In European Radiology, a prospective crossover study across five European centres, led by Russo, compared a FIGO-anchored structured MRI report with narrative reporting for staging cervical cancer in 100 patients [6]. Gynaecologists and radiation oncologists rated both formats. Ninety-eight percent of structured reports met the satisfaction threshold, against 78 percent of narrative reports. The structured format scored higher on every general item, with the largest gains in readability and brevity. The advantage held across centres and stages. The endpoint is clinician satisfaction, so staging accuracy and patient outcomes were not tested.

In Radiology, Liu and colleagues scanned 30 patients with multiple sclerosis from a Chinese registry at both 7 tesla and 3 tesla on the same day [7]. At 7 tesla, readers counted about 28 percent more lesions overall, and more than twice as many cortical lesions. The share of lesions with a central vein sign roughly doubled, to about two thirds. The proportion of paramagnetic rim lesions was similar at both field strengths. Agreement between readers was excellent. The sample is small and every patient already had a diagnosis, so the study does not show whether 7 tesla changes diagnosis in suspected disease.

Three quick signals.

In European Radiology, Ji and colleagues compared photon-counting and conventional CT for gastric cancer T staging in two separate retrospective groups of about fifty patients each [8]. Photon-counting CT gave higher signal and contrast to noise, left fewer lesions unidentified, and showed numerically higher staging accuracy. The authors call it exploratory and state that validation is still needed.

A scoping review in European Radiology by Pakpoor and colleagues covered 21 studies of seven commercial AI devices for prostate MRI [9]. Most of these studies were retrospective and single-centre. Gains were inconsistent, appeared larger for less-experienced readers, and sometimes traded sensitivity for specificity. No study assessed cost-effectiveness or patient outcomes.

In a single-centre retrospective series of 58 patients in European Radiology, Mathy and colleagues found that stent-grafting for hepatic artery bleeding or pseudoaneurysm was technically successful in 97 percent [10]. Primary clinical success was 88 percent. Four millimetre grafts carried roughly eight times the occlusion risk, and patency fell to just over half at two years.

Stepping back across this week's papers, most of the evidence rests on intermediate endpoints: management changes, clinician satisfaction, lesion counts, image quality metrics and tissue concentrations. Very little of it reaches patient outcomes. Many of the studies are small, single-centre or retrospective, and several apply criteria in hindsight. The pooled analyses are broader, but they inherit the limits of the studies beneath them, so most of this week's signals still await prospective confirmation.

If you only have time for one paper this week, make it the Lung-RADS 2022 analysis of cancers associated with cystic airspaces [1]. Drawing on three screening trials, it puts a size on a pattern that is easy to dismiss as benign, and it tests whether the newest cyst criteria would have caught these cancers sooner.

That's your roundup for This Week in Radiology. The full transcript and references are available on the episode page. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And a quick rating on Apple Podcasts or Spotify helps other physicians discover the show.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.