RadioRAG uses online retrieval–augmented generation to potentially improve the accuracy and factuality of large language model–generated responses to case-based radiology questions by incorporating real-time data from Radiopaedia.

Visual large language models (VLLMs) are discussed as potential tools for assisting radiologists in image interpretation, yet their clinical value remains unclear. This study provides a systematic and comprehensive comparison of general-purpose and biomedical VLLMs in radiology. We evaluated 180 representative clinical images with validated reference diagnoses (radiography, CT, MRI; 60 each) using seven VLLMs (ChatGPT-4o, Gemini 2.0, Claude Sonnet 3.7, Perplexity AI, Google Vision AI, LLaVA-1.6, LLaVA-Med-v1.5). Each model interpreted the image without and with clinical context. Mixed-effects logistic regression models assessed the influence of model, modality, and context on diagnostic performance and hallucinations (fabricated findings or misidentifications). Diagnostic accuracy varied significantly across all dimensions (p ≤ 0.001), ranging from 8.1% to 29.2% across models, with Gemini 2.0 performing best and LLaVA performing weakest. CT achieved the best overall accuracy (20.7%), followed by radiography (17.3%) and MRI (13.9%). Clinical context improved accuracy from 10.6% to 24.0% (p < 0.001) but shifted the model to rely more on textual information. Hallucinations were frequent (74.4% overall) and model-dependent (51.7–82.8% across models; p ≤ 0.004). Current VLLMs remain diagnostically unreliable, heavily context-biased, and prone to generating false findings, which limits their clinical suitability. Domain-specific training and rigorous validation are required before clinical integration can be considered.

OpenAI’s GPT-4 with Advanced Data Analysis autonomously performed basic statistical analyses and built machine learning–based models to evaluate a large chest radiography dataset, performing similarly to its manually crafted counterpart.

Background and Aims: Large language models (LLMs) can potentially support clinicians in their daily routine by providing easy access to information. Yet, they are plagued by stating incorrect facts and hallucinating when queried. Increasing the context by providing external databases while prompting LLMs may decrease the risk of misinformation. This study compares the influence of increased context on the coherence of LLM‐based treatment recommendations with the recently updated WHO guidelines for the treatment of chronic hepatitis B (CHB). Methods: GPT‐4 was queried with five clinical case vignettes in two configurations: with and without additional context. The clinical vignettes were explicitly constructed so that treatment recommendations differed between the formerly applicable 2015 WHO guidelines and the updated 2024 ones. GPT‐4 with context was provided access to the updated guidelines, while GPT‐4 without context had to rely on its internal knowledge. GPT‐4 was accessed only a few days after the release of the new WHO guidelines. Treatment recommendations were compared regarding guideline coherence, information inclusion, textual errors, wording clarity and preciseness by seven physicians. Results: Using GPT‐4 with context increased the coherence of the treatment recommendations with the new 2024 guidelines from 51% to 91% compared to GPT‐4 without context. Similar trends were observed for all other categories, leading to an increase of 54% in preciseness and clarity, 24% in completeness of incorporating the case vignette information, and 12% in textual correctness. Conclusions: If LLMs are consulted by clinicians for medical advice, they should be given access to external data sources to increase the chance of providing factually correct advice.

Introduction Aim of this study is to evaluate whether magnetic particle imaging (MPI) is capable of measuring velocities occurring in the coronary arteries and to compute coronary flow reserve (CFR) in a canonical phantom as a preliminary study. Methods For basic velocity measurements, a circulation phantom was designed containing replaceable glass tubes with three varying inner diameters, matching coronary-vessel diameters. Standardised boluses of superparamagnetic-iron-oxide-nanoparticles were injected and visualised by MPI. Two image-based techniques were competitively applied to calibrate the respective glass tube and to compute the mean velocity: full-duration-at-half-maximum (FDHM) and tracer dilution (TD) method. For CFR-calculation, four necessary settings of the circulation model of a virtual vessel with an inner diameter of 4 mm were generated using differently sized glass tubes and a stenosis model. The respective velocities in stenotic glass tubes were computed without recalibration. Results On velocity level, comparison showed a good agreement (r(FDHM) = 0.869, r(TD) = 0.796) between techniques, preferably better for 4 mm and 6 mm inner diameter glass tubes. On CFR level MPI-derived CFR-prediction performed considerably inferior with a relative error of 20-44%. Conclusions MPI has the ability to reliably measure coronary blood velocities at rest as well as under hyperaemia and therefore may be suitable for CFR calculation. Calibration-associated accuracy of CFR-measurements has to be improved substantially in further studies.

Background: Sarcopenia assessed by skeletal muscle area (SMA) at the third lumbar vertebra (L3) is an established prognostic marker in many malignancies, including head and neck cancer (HNC). However, in HNC, L3 is rarely assessed. The prognostic value of myosteatosis, measured by skeletal muscle radiation attenuation (SMRA) remains largely unexplored. This study evaluated both muscle metrics at the third cervical vertebra (C3) for locoregional control (LRC) and overall survival (OS) in HNC. Methods: SMA and SMRA at C3 were quantified in CT scans of 904 HNC cases by a deep learning-based segmentation pipeline with manual verification. Cox proportional hazards models assessed associations with LRC and OS. Results: Median SMA was 36.64 cm² (IQR: 30.12–42.44). Median SMRA was 50.77 HU (IQR: 43.04–57.39). In multivariable analysis, lower SMA (HR 1.85, 95% CI: 1.19–2.88, p ≤ 0.001) and lower SMRA (HR 1.76, 95% CI: 1.22–2.54, p < 0.001) were associated with lower LRC. For OS, lower SMA (HR 1.53, 95% CI:1.06–2.20, p = 0.02) and lower SMRA (HR 2.13, 95% CI: 1.58–2.88, p < 0.001) were associated with a worse outcome in multivariable analysis. Conclusions: Both SMRA and SMA assessed at C3 correlate with worse LRC and OS in HNC.