This Week in Radiology — Oct 3, 2026

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

Welcome to This Week in Radiology. This week we're covering 10 notable papers spanning oncologic and abdominal imaging, dose reduction and interventional technique, and quantitative neuro and musculoskeletal MRI alongside a look at department workflow. Let's dive in.

Pancreatic imaging leads this week, with two papers in European Radiology that address different but overlapping diagnostic questions. Rezuș and colleagues pooled forty direct comparative studies in a systematic review and meta-analysis, each enrolling patients suspected of pancreatic ductal adenocarcinoma who underwent at least two imaging methods [1]. For simply detecting the tumour, CT, MRI and endoscopic ultrasound performed without meaningful difference. The modalities diverged on staging. MRI picked up liver metastases in close to nine in ten cases, against roughly six in ten for CT, and was also more specific. CT, in turn, was better for vascular invasion and more specific for resectability, at just under ninety percent compared with about sixty percent for MRI. Endoscopic ultrasound was the most sensitive test for nodal disease. The authors read this as complementary roles, with liver MRI adding value before potentially curative surgery to exclude small metastases and CT remaining the anchor for vascular and resectability assessment. Because this is pooled data from heterogeneous studies, it supports rather than settles a combined pathway. Staying in the pancreas, Nagayama and colleagues tackled a familiar diagnostic trap, the intrapancreatic accessory spleen masquerading as a hypervascular neoplasm when only a portal venous phase is available [2]. In 85 patients, ratios comparing lesion to spleen on iodine maps, effective atomic number maps and low energy monochromatic images separated the two entities far better than conventional images, with discrimination approaching near perfect for tail lesions. When three radiologists added spectral images to their visual read, accuracy rose by more than a quarter, to nearly ninety percent, and agreement between readers improved substantially. This is a single retrospective cohort, but it suggests that spectral data already sitting in a routine scan may resolve lesions that would otherwise go on to further workup.

Two more oncology papers look at how imaging stratifies risk and prevents cancer. Tsukamoto and colleagues, also in European Radiology, examined human papillomavirus associated oropharyngeal carcinoma, where clinical and pathological nodal staging under the ninth edition classification often disagree [3]. Counting imaging-detected nodal metastases, defined by FDG PET uptake or a Node-RADS score of three or higher, they found that patients with five or more such nodes had roughly an elevenfold higher risk of death and about a sixfold higher risk of recurrence or disease-related death, and the association held after adjusting for clinical stage. With only 68 patients the estimates are imprecise, so this is a hypothesis-generating signal that node count could mirror pathological staging, not yet a staging rule. Meanwhile in RadioGraphics, Taya, Pickhardt and colleagues review where CT colonography stands now that Medicare coverage began in 2025, ending a fifteen-year policy stalemate [4]. They argue that CT colonography matches colonoscopy for clinically significant polyps, carries an excellent safety record, and alongside colonoscopy is one of only two tests accurate enough for primary prevention, whereas stool and blood tests are largely limited to cancer detection. They emphasise that cathartic preparation, stool tagging and good distention are essential for flat serrated lesions, and that selective polypectomy avoids removing pseudodisease. This is an expert narrative review rather than new data, and the authors themselves flag worsening underutilisation driven by practice and referral barriers.

Our second theme is radiation and contrast dose, and the procedures where imaging guides intervention. In the European Journal of Radiology, Liu and colleagues randomised 70 patients with a body mass index under 28 to standard CT pulmonary angiography or an ultra triple-low protocol at 60 kilovolts, using just 10 millilitres of contrast at a slower injection rate [5]. The low protocol cut effective dose by about 85 percent and contrast volume by two thirds, and when paired with a deep learning reconstruction, pulmonary arteries all exceeded 250 Hounsfield units with better noise and subjective scores than standard reconstructions. Image quality is not diagnostic accuracy for embolism, and slimmer patients only were studied, so the authors appropriately call for further validation. A different dose question appears in European Radiology from Schneider and colleagues, who compared dual-phase angio-CT with single-phase cone beam CT across 92 intra-arterial liver procedures [6]. The three-dimensional imaging dose was about four times lower with angio-CT, and total dose looked lower too, but that total-dose difference lost statistical significance once conversion uncertainty was accounted for. The fair reading, and the authors' own conclusion, is that angio-CT does not increase patient exposure, rather than that it reduces it. In the American Journal of Neuroradiology, Sakamoto and colleagues followed 82 patients with skull base meningiomas embolised before surgery [7]. Tumours shrank by about ten percent on average within roughly four days, and volume fell in about ninety percent of patients, though responses ranged widely, with some tumours actually growing. Greater loss of intratumoral enhancement on post-embolisation MRI tracked with greater shrinkage. Permanent neurological deficits occurred in two patients, a reminder that the single-centre retrospective data do not resolve whether the surgical benefit outweighs procedural risk.

Our final theme covers quantitative MRI and the practice environment around it. The American Journal of Roentgenology published an expert panel review from Cogswell, Barkhof and colleagues on idiopathic normal pressure hydrocephalus [8]. The panel recommends an unenhanced brain MRI including a three-dimensional T1-weighted sequence, and identifies ventriculomegaly with narrowed high convexity sulci and enlarged Sylvian fissures, the pattern known as DESH, together with aqueduct patency as the key supportive features. For incidental cases, they offer standard impression wording, findings supportive of a diagnosis of normal pressure hydrocephalus in the appropriate clinical context, while acknowledging that the prognostic value of these features has varied across studies. In the Journal of Magnetic Resonance Imaging, Raya and colleagues report a prospective cohort of 71 participants testing diffusion tensor imaging of knee cartilage [9]. Mean diffusivity in the medial femoral condyle distinguished knees with the earliest radiographic osteoarthritis from normal knees very well, while a T2 mapping model performed little better than chance. Among knees with minimal baseline change, close to a third of those knees progressed over three years, and adding diffusivity improved prediction of joint space narrowing. Follow-up was available in only 40 participants and the technique is specialised, so this is early evidence for a research biomarker. Finally, in the Journal of the American College of Radiology, Sreedharan and colleagues describe the REDUCE-CALL study [10]. After introducing chat messaging within the electronic record plus targeted education, daily phone calls to the duty radiologist more than halved, to about 80 a day, total communication fell substantially, protocolling volume was unchanged, and chat messages received a reply within a median of about a quarter of an hour. It is a single-site before and after comparison, but the effect size is large.

If you only have time for one paper this week, make it the pancreatic cancer meta-analysis from Rezuș and colleagues in European Radiology [1]. It reframes the CT versus MRI debate in pancreatic cancer as a question of which staging task each modality does best, which bears directly on how preoperative pathways are built.

Here is what this week's evidence adds up to in Radiology. First, pooled comparative data support CT and MRI as complementary in pancreatic cancer, with MRI stronger for liver metastases and CT stronger for vascular involvement and resectability, though study heterogeneity limits certainty. Second, spectral CT and very low kilovolt protocols with deep learning reconstruction show promising single-centre results, for characterising accessory spleen and for cutting dose and contrast in pulmonary angiography, but neither has outcome or accuracy validation at scale. Third, imaging-based nodal counts in oropharyngeal cancer and cartilage diffusion metrics in early osteoarthritis are emerging prognostic biomarkers built on small cohorts that need external confirmation. Fourth, angio-CT appears not to raise patient dose in liver interventions, and preoperative meningioma embolisation produces fast but highly variable shrinkage, leaving its net benefit unsettled. And fifth, expert reviews on normal pressure hydrocephalus and CT colonography, plus a large workflow effect from asynchronous chat, point to practice and policy levers that sit largely within radiology's control.

That's your roundup for This Week in Radiology. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And if it saves you time, a quick rating in your podcast app helps other physicians find the show.

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