To evaluate the impact of night shift-induced fatigue on radiology residents’ diagnostic performance and to assess whether AI assistance offsets potential performance loss. In this prospective intra-individual reader study, ten radiology residents (mean experience: 35 months; three women) interpreted three sets of 33 chest radiographs before and after a first-call on-site night shift. Sessions included two pre-shift readings (morning/afternoon, without AI) and two post-shift readings (morning, with/without AI). Eye tracking captured fixation duration and saccade count, and fatigue was defined as a ≥ 5
Emergency departments face increasing overcrowding, resulting in diagnostic delays, higher mortality, and rising healthcare costs. Early risk stratification is essential to optimise clinical decision-making. Although medical imaging plays a central role in initial assessment, workload pressures and operational constraints have driven the development of artificial intelligence (AI) tools to support triage and prioritisation in emergency radiology. To map and synthesise evidence on the applications, performance, and limitations of AI in emergency radiology for early risk stratification in acute care. A scoping review was conducted in accordance with PRISMA-ScR guidelines and Joanna Briggs Institute recommendations. The protocol was developed a priori and registered in the Open Science Framework. A systematic search of PubMed, Scopus, and Web of Science identified studies published between 2015 and 2025 that evaluated AI applications in medical imaging in acute care settings. Evidence was synthesised using conceptual mapping and thematic analysis. Seventy-one studies were included. Computed tomography was the most frequently evaluated modality, followed by radiography and ultrasound. Major domains included neurovascular disease, acute cardiopulmonary conditions, trauma, COVID-19, and acute aortic syndromes. Studies reported high diagnostic performance and potential workflow benefits; however, evidence was predominantly retrospective and internally validated, with limited prospective evaluation, external validation, or evidence of improved patient-level outcomes. AI applications in emergency radiology show promising diagnostic and workflow-support capabilities. However, current evidence largely reflects early-stage development, with limited prospective validation, scarce real-world impact assessment, and insufficient evaluation of clinically meaningful outcomes. Question How clinically mature are AI applications designed to support early risk stratification and prioritisation in time-sensitive emergency radiology workflows? Findings Most studies report high diagnostic performance in time-sensitive conditions, yet evidence remains predominantly retrospective with limited external validation or prospective real-world implementation. Clinical relevance AI may support earlier identification and prioritisation of critical imaging findings, but routine adoption requires prospective external validation and evidence of meaningful effects on clinical decisions and patient outcomes.
The presence of extra-pancreatic nerve plexus invasion (EPNI) in pancreatic ductal adenocarcinoma (PDAC) is known to correlate with aggressive disease, however, it is unknown whether specific pathways or morphologies carry different prognostic implications. This study evaluates whether distinct CT-defined patterns of EPNI predict mortality in PDAC. This retrospective, single-center study included 89 patients with pathologically confirmed PDAC who underwent preoperative pancreatic protocol multidetector CT (MDCT) followed by surgical resection. Patients were identified from an institutional database and required available preoperative CT and histopathology. Two blinded, experienced radiologists reviewed MDCT images to assess EPNI, which was categorized by anatomic pathway and morphologic pattern. Overall survival was compared between EPNI-positive and EPNI-negative groups and among EPNI subtypes using log-rank testing, with stratification by tumor stage. Cox proportional hazards models adjusted for age, sex, and surgical margin were performed. A total of 89 patients (M/F: 54/35), median age: 66 years (IQR, 60–70 years), were evaluated. CT evidence of EPNI was present in 65 patients (73
Yttrium-90 transarterial radioembolization (TARE) combined with immune checkpoint inhibitors has been investigated as a potential therapeutic strategy for advanced hepatocellular carcinoma (HCC); however, evidence supporting its addition to current standard systemic therapy remains limited. We aimed to compare clinical outcomes in patients with advanced HCC treated with TARE followed by atezolizumab plus bevacizumab (TARE–Atezo/Bev) and Atezo/Bev alone. This retrospective cohort study included systemic therapy-naïve patients with advanced HCC treated with Atezo/Bev alone or TARE–Atezo/Bev at a single tertiary hospital between 2018 and 2025. Baseline characteristics were adjusted using propensity score matching (PSM) and inverse probability of treatment weighting (IPTW). The primary outcome was overall survival (OS), and the secondary outcome was progression-free survival (PFS). A total of 156 eligible patients were enrolled, including 93 in the Atezo/Bev group and 63 in the TARE–Atezo/Bev group. In the unmatched cohort, median OS was significantly longer in the TARE–Atezo/Bev group than in the Atezo/Bev group (14.1 vs 6.4 months; hazard ratio=0.43, 95
