To synthesise evidence on the diagnostic performance, safety, feasibility and clinical impact of ultra-low-field portable magnetic resonance imaging (pMRI) in adult neurocritical care and intensive care unit (ICU) populations. PubMed, Scopus and Embase were searched from inception to 6 February 2026 and supplemented by grey literature and citation searching. We included original English-language studies of adult ICU, neurocritical care or emergency department patients undergoing pMRI at ≤ 0.1 T. Two reviewers independently screened studies and extracted data. Risk of bias in diagnostic comparisons was assessed with an adapted QUADAS-2 framework, and feasibility and safety outcomes were appraised narratively. Because of heterogeneity, findings were synthesised narratively. The review was prospectively registered with PROSPERO (CRD420261302014). Eleven studies published between 2020 and 2024 were included, comprising 694 patient participants and 715 patient examinations, plus 40 healthy controls. Cohorts overlapped substantially: six studies came wholly and one partly from a single institution, and one ECMO study was a subset of another. The number of unique patients is therefore estimated at 281 to 677. In the multicentre ECMO study, adverse events occurred in 3 of 50 patients (6
Radiographs are the first line imaging modality in evaluation for fractures, which can be challenging in equivocal cases with findings that mimic fractures. Accurate and quick identification of acute fractures is essential for triage, management, and prevention of complications. At the same time, overcalling a benign finding as a fracture can lead to patient anxiety and further testing that adds health care cost and time. As such, familiarity with fracture mimics is a critical skill for clinicians involved in emergency and acute care settings. This pictorial essay reviews common acute fracture patterns and highlights key imaging features that distinguish acute fractures from benign mimics, primarily through numerous representative cases. The article also provides a framework to address equivocal findings with next steps that can be performed for troubleshooting.
General-purpose multimodal large language models (LLMs) are being applied to radiology, but their performance under the simplest realistic clinical query (a single binary fracture-or-not prompt) has not been rigorously quantified. We evaluated ChatGPT-5 on this task using a public distal radius dataset. In this retrospective diagnostic accuracy study, 396 anteroposterior distal radius radiographs from the Stanford MURA dataset were submitted to ChatGPT-5, each in a fresh temporary chat session. The prompt was: "Is there a fracture in this X-ray? Answer with yes or no." The reference standard was an image-level label assigned by consensus between a senior orthopedic resident and a board-certified orthopedic surgeon; MURA study-level labels were used only to select candidate images. Twenty-six radiographs (6.6
