Retrieval-Augmented Generation Elevates Local LLM Quality in Radiology Contrast-Media Consultation

Retrieval-Augmented Generation Elevates Local LLM Quality in Radiology Contrast-Media Consultation | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Retrieval-Augmented Generation Elevates Local LLM Quality in Radiology Contrast-Media Consultation Akihiko Wada, Yuya Tanaka, Mitsuo Nishizawa, Akira Yamamoto, Toshiaki Akashi, and 8 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6538325/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 02 Jul, 2025 Read the published version in npj Digital Medicine → Version 1 posted 9 You are reading this latest preprint version Abstract Large language models (LLMs) excel at text reasoning, but hospital use is curtailed by privacy rules and limited medical tuning. We asked whether retrieval-augmented generation (RAG) can lift a lightweight, locally deployable LLM to near-cloud performance. In 100 simulated iodinated-contrast-media consultations we compared Llama 3.2-11B (baseline and RAG) with three cloud-based models—GPT-4o mini, Gemini 2.0 Flash and Claude 3.5 Haiku. A blinded radiologist ranked the five replies per case, and three LLM judges scored accuracy, safety, structure, tone, applicability and latency. RAG eliminated hallucinations (0% vs 8%; χ²₍Yates₎ = 6.38, p = 0.012) and improved mean rank by 1.3 (Z = − 4.82, p < 0.001), narrowing most of the quality gap while remaining faster (2.6 s vs 4.9–7.3 s). The LLM judges, who weighted latency, preferred the RAG-enhanced Llama, whereas the radiologist ranked GPT-4o mini third. RAG thus enables near-cloud performance on-premise without exposing patient data. Health sciences/Health care/Medical imaging Health sciences/Health care Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Large language models (LLMs) have rapidly expanded across domains, demonstrating significant potential to enhance human capabilities and workflow efficiency in specialized fields that require domain-specific knowledge. In medicine, LLMs are promising to summarize medical literature, support clinical reasoning, and extract key information from medical records 1,2 . In radiology, recent studies highlight LLMs' potential to improve report quality, workflow efficiency, and diagnostic accuracy 3,4 . Despite this potential, healthcare implementation of LLMs faces unique challenges centered on a fundamental dilemma: balancing patient data confidentiality with advanced AI capabilities. Cloud-based LLMs offer superior reasoning abilities through large-scale training data but require transmitting sensitive medical information to external servers, raising substantial concerns under regulations like HIPAA and GDPR 5 . Conversely, locally deployed LLMs ensure data privacy but typically demonstrate inferior performance due to constraints in model size and computational resources 6 . This performance gap is particularly pronounced in specialized medical domains where domain-specific knowledge and nuanced understanding of clinical guidelines are essential. Retrieval-augmented generation (RAG) has emerged as a promising approach to address this trade-off. By integrating external domain-specific knowledge into the response generation process, RAG enables LLMs to leverage specialized information beyond their training data, reduce hallucinations through authoritative source grounding, and dynamically update knowledge without additional model fine-tuning 7 . RAG combines high accuracy with robust data privacy when implemented within a locally deployed environment, potentially offering a solution to the privacy-performance dilemma in healthcare AI. While RAG's effectiveness has been demonstrated in general knowledge retrieval tasks, its utility in highly specialized medical fields, particularly radiology, remains underexplored. Previous studies have identified knowledge base quality and retrieval efficiency as key determinants of RAG performance in medical applications 8,9 , but few have evaluated its application in time-sensitive clinical decision support scenarios 10 . To systematically assess locally deployed RAG-enhanced LLMs in a specialized medical context, this study focuses on iodinated contrast media (ICM) consultations—a crucial radiological task requiring specialized knowledge and real-time decision-making. CT examinations have become fundamental to modern diagnostic imaging, with annual scan volumes in the United States increasing from approximately 3 million in 1980 to 91 million in 2019 11 . Notably, 40–48% of all CT scans involve ICM to enhance visualization of vascular and organ structures, making appropriate contrast administration essential to radiological practice. Recent challenges have amplified the importance of precise ICM consultation. The COVID-19 pandemic caused severe global contrast agent shortages, necessitating careful assessment of contrast-enhanced scan appropriateness 12 . Additionally, ICM-related adverse events range from mild allergic reactions to severe nephrotoxicity, making risk stratification crucial for patient safety. Research indicates that 7.5% of ICM-related complications could be prevented through appropriate consultation 13 , while communication errors in contrast media ordering account for 13.9% of radiology workflow inefficiencies 14 . Effective ICM consultation requires specialized radiological knowledge, clinical experience, and dynamic risk assessment capabilities. LLMs could potentially assist radiologists by automating risk assessment and protocol selection, thereby reducing workload and enhancing decision-making. The integration of facility-specific protocols and recently updated guidelines through RAG could be particularly valuable, as contrast media administration often involves institution-specific rules that cannot be captured in general model training. This study aims to evaluate the performance of RAG-enhanced lightweight local LLMs in ICM consultation compared to leading cloud-based models. We developed and assessed a RAG-implemented locally deployable model (Llama 3.2 11B) against three cloud-based LLMs (GPT-4o mini, Gemini 2.0 Flash, Claude 3.5 Haiku) using 100 simulated clinical scenarios requiring ICM risk assessment and protocol selection. Both radiologists and automated LLM evaluators evaluated performance through a comprehensive assessment, examining clinical accuracy, safety, response structure, communication quality, practical applicability, and response time. By quantifying performance differences between cloud-based and locally deployed models and assessing whether RAG can substantially enhance locally deployable model performance while maintaining data privacy, this research provides practical evidence to guide healthcare institutions in selecting and implementing AI systems for clinical decision support. Methods 1. Consultation Scenarios Development We designed 100 simulated consultation scenarios covering a diverse range of iodinated contrast media (ICM) use cases, including appropriateness, contraindications, incomplete orders, and avoidable contrast use. These scenarios systematically reflected five key categories in radiological practice: (1) appropriateness consultations for specific clinical indications, (2) optimal contrast agent selection and protocols, (3) contraindication identification and risk assessment, (4) recognition of incomplete ordering information, and (5) identification of cases where contrast could be avoided. Scenarios were initially generated using GPT-4o and Claude 3.5 Sonnet with structured prompts for clinical authenticity. To ensure realism and guideline compliance, two radiologists (4 and 30 years of experience, respectively) independently reviewed and revised each scenario according to current ACR and ESUR guidelines. Special attention was given to incorporating realistic clinical nuances and decision-making challenges commonly encountered in ICM consultations. The complete set of scenarios is available in the supplemental materials. 2. Language Models Configuration We benchmarked five LLMs—three cloud-based models (GPT-4o mini from OpenAI, Gemini 2.0 Flash from Google, and Claude 3.5 Haiku from Anthropic) accessed through their public REST APIs, and two locally deployable models (Llama 3.2 11B from Meta, as a baseline and RAG-enhanced variant). Llama 3.2 11B is designed as a locally deployable, lightweight parameter model ideal for on-premises computing environments. An example of the local execution environment for the Llama 3.2 11B model is as follows: NVIDIA RTX 3090 GPU (24 GB VRAM; Graphics Processing Unit), Intel Core i7-12700K CPU (8 cores; Central Processing Unit), and 32 GB RAM (Random Access Memory). This study deployed the Llama 3.2 11B via GroqCloud's hosted inference API to facilitate direct performance comparisons with the cloud-native models. All models were deployed via the Dify platform (version 0.10.0) with standardized parameters, including consistent temperature settings (0.2) and single-shot response generation 15 . For each model, we recorded both the responses and the response times while maintaining all other default settings for consistent comparison. For the automated evaluation process, we employed three high-performance models (GPT-4o, Gemini 2.0 Flash Thinking Experimental, and Claude 3.5 Sonnet) as evaluators. Table 1 documents the technical specifications for all models used in this study, including parameter sizes, deployment characteristics, and implementation details. 3. Retrieval-Augmented Generation Implementation For the RAG-enhanced model, we compiled a knowledge base from authoritative sources, including the ACR Manual on Contrast Media, ESUR guidelines, institutional protocols, and relevant literature 16–18 . Knowledge entries were initially generated in a question-answer format using Claude 3.5 Sonnet and were subsequently reviewed, modified, and organized by two radiologists (Table 2). The knowledge base was structured as segments with a maximum of 600 tokens, comprising 66 chunks with an average of 337 characters per segment. These segments were transformed into high-dimensional embeddings using OpenAI's text-embedding-3-large model and indexed in "High Quality" mode. The system employed a hybrid search method that combined semantic vector search with keyword-based retrieval. For each clinical query, the system retrieved four relevant context fragments (TopK = 4) based on cosine similarity calculations. These retrieved fragments were then ranked according to their relevance scores and incorporated into the prompt structure for the Llama 3.2 11B model (Fig. 2 ). 4. Evaluation Methodology We used a two-tier evaluation strategy: (i) human expert review by a board-certified radiologist, and (ii) automated scoring by three large LLM “judges.” For the human tier, the radiologist reviewed all responses in a blinded manner, ranking the anonymized responses from 1st (best) to 5th (worst) based on clinical appropriateness and practical applicability, without considering response time. In addition to rubric-based scoring, each response was screened for hallucinations, defined as clinically incorrect or guideline-inconsistent statements that could alter patient management (e.g., wrong dosage, missing contraindication). Hallucination presence (yes/no) was recorded per scenario, and the frequencies were compared between models using χ² (with Yates correction) and Fisher’s exact tests. For automated LLM “judge” evaluation, we utilized three high-performance LLM evaluators within a G-E VAL-like framework 19 . Each response was systematically compared against reference standards developed by radiology experts, using six predefined criteria: Clinical Accuracy: Alignment with established medical guidelines and factual correctness Safety: Proper identification of contraindications and emphasis on patient safety measures Response Structure: Clarity and logical organization of the answer Professional Communication: Appropriate use of medical terminology and professional language Practical Applicability: Provision of actionable clinical recommendations Response Time: Performance against predefined speed thresholds LLM evaluators assessed the responses based on these criteria, with final rankings determined by the total combined score. In cases of tied scores, models were assigned the same higher rank. Comprehensive details regarding the evaluation prompts and scoring rubrics are available in the Supplementary Materials. 5. Statistical Analysis All analyses were conducted in Python 3.9.13 with scipy 1.10.0 and statsmodels 0.13.5. Because the Shapiro–Wilk test revealed significant departures from normality (p < 0.05) in the rank distributions, we exclusively applied non-parametric procedures. Overall differences among the five LLMs were examined with the Friedman test, which accounts for the repeated-measures nature of the 100 shared consultation scenarios. Pairwise contrasts were then assessed with the Nemenyi post-hoc test; for the two comparisons of primary interest—GPT-4o mini versus the RAG-enhanced Llama, and baseline versus RAG Llama—we also report Mann-Whitney U (Wilcoxon rank-sum) statistics with Bonferroni-adjusted significance thresholds. Effect sizes are expressed as the difference in mean ranks (Δ), accompanied by Wilcoxon Z values and two-sided p values. Descriptively, we present mean rank ± standard deviation and 95% confidence intervals to visualise performance spread and inter-evaluator agreement, following conventions established in recent LLM benchmarks such as MT-Bench and Chatbot Arena. 6. Ethical Considerations This study did not require Institutional Review Board approval as it utilized simulated clinical scenarios rather than actual patient data. All consultations were fictional cases designed to reflect common clinical situations without incorporating any protected health information. Results 1. Overall Model Performance and Rankings A global Friedman test across the five LLMs confirmed significant differences in median ranks for the 100 iodinated-contrast-media scenarios (χ² = 78.4, p < 0.001). Post-hoc Nemenyi comparisons showed that the three cloud models clustered at the top: Gemini 2.0 Flash was dominant (radiologist mean rank = 1.36; first in 74% of cases), followed by Claude 3.5 Haiku (2.09) and GPT-4o mini (3.21). The baseline Llama 3.2 11B was ranked last in 86% of cases (mean 4.80). RAG lifted its performance to a mean rank of 3.54, close to GPT-4o mini’s 3.21 (Δ = − 1.26; Z = − 7.26, p < 0.001). Automated LLM evaluators confirmed the overall cloud supremacy, again ranking Gemini 2.0 Flash and Claude 3.5 Haiku first and second, respectively. Within the remaining models, however, they placed the RAG-enhanced Llama 3.2 11B as 3rd, ahead of GPT-4o mini (Δ = − 0.44 to − 0.82; Z = + 2.56 to + 5.43; p ≤ 0.009). This reversal relative to the radiologist’s preference illustrates a systematic divergence between human and LLM-based judgements. 2. Inter-Evaluator Agreement and Assessment Patterns In three LLM evaluators, Claude 3.5 Sonnet exhibited the highest agreement with the radiologist at 82.2%, suggesting its assessments most closely align with human expert judgment. GPT-4o’s evaluation also showed a relatively high agreement rate of 75.5% compared to the radiologist. Across all evaluators, there was broad consensus on the top performer (Gemini 2.0 Flash, 81.6% agreement) and bottom performer (base Llama 3.2 11B), while mid-tier models showed more variable assessments. The most significant evaluation divergence appeared in mid-tier rankings. All three LLM evaluators consistently ranked the RAG-enhanced Llama 3.2 11B higher than GPT-4o mini, contrary to the radiologist's assessment. This divergence was most pronounced with GPT-4o, which ranked the RAG-enhanced Llama second (mean rank: 2.53) and GPT-4o mini fourth (mean rank: 3.35). 3. Clinical Quality and Hallucination Mitigation Qualitative assessment by a radiologist revealed that hallucinations—particularly in contrast dosing and contraindication recommendations—were present in 8% of responses from the base Llama model and were eliminated (0%) in the RAG-enhanced version (χ²₍Yates₎ = 6.38, p = 0.012; Fisher p = 0.0068). These hallucinations typically involved incorrect ICM protocols, including improper contraindication identification and dosage recommendations, which could potentially impact patient safety. None of the cloud-based models exhibited hallucinations in the evaluated scenarios. Among the 100 cases, 54% showed marked improvement with RAG—most notably in safety-critical content such as contraindication management and precise dosing guidance—while an additional 33% demonstrated minor gains involving non-critical factual details. These improvements were concentrated in cases that required specialized knowledge of institutional protocols or recently updated guidelines. 4. Multidimensional Performance Assessment Radar chart analysis of LLM evaluations indicated substantial gains in clinical accuracy, safety, and applicability for the RAG-enhanced model, with modest changes in communication and structure (Fig. 4 ). The most significant improvements occurred in domains directly related to patient safety and clinical decision-making. Specifically, clinical accuracy scores improved from a mean of 14.2 to 21.5 points (out of 25), and safety considerations improved from 11.8 to 18.7 points (out of 20). Professional communication showed moderate enhancement (13.4 to 16.8 out of 20), while response structure remained relatively stable (7.8 to 8.3 out of 10), as the base model already demonstrated adequate formatting capabilities. 5. Response Time Analysis In terms of response time, the RAG-enhanced model (2.58s) remained faster on average than all cloud-based models (GPT-4o mini: 4.87s, Claude 3.5 Haiku: 7.25s, Gemini 2.0 Flash: 3.14s), despite increased latency compared to its base version (1.31s) (Fig. 5 ). The difference was statistically significant for all comparisons, including Gemini 2.0 Flash (t = − 3.96, p = 0.00011). Response time variability differed notably across models. Claude 3.5 Haiku demonstrated the most consistent performance with a coefficient of variation (CV) of 10.76%, while the RAG-enhanced Llama 3.2 11B showed the highest variability (CV = 44.19%), likely due to fluctuations in knowledge retrieval latency. The base Llama 3.2 11B (CV = 37.40%), GPT-4o mini (CV = 31.21%), and Gemini 2.0 Flash (CV = 26.43%) exhibited intermediate variability. While the RAG-enhanced model's response time was more variable, the trade-off was acceptable given the substantial performance improvements in clinical accuracy and safety. Discussion RAG Technology in Radiological Applications This study demonstrates that a retrieval-augmented local LLM can achieve clinically reliable performance in a safety-critical radiological use case while preserving patient privacy. By integrating curated guideline-based knowledge into a lightweight LLM, we observed significant improvements in clinical accuracy, risk stratification, and communication quality—criteria that are central to iodinated contrast media (ICM) decision-making. Notably, the RAG-enhanced model eliminated hallucinations observed in the base model and achieved faster average response times than all evaluated cloud-based LLMs, suggesting feasibility for real-time use in clinical settings. Our findings align with recent research demonstrating RAG's potential to enhance LLM performance in specialized medical domains. Gupta et al. (2024) emphasized RAG's application in medical knowledge retrieval, showing improvements in accuracy and reliability 20 . Mortaheb et al. (2025) introduced a multimodal evaluation framework ensuring contextual relevance 9 , while Nguyen et al. (2024) reported significant accuracy improvements in clinical question-answering when incorporating RAG into language models 8 . These studies collectively confirm that integrating domain-specific knowledge improves evidence-based reasoning capabilities of LLMs, leading to enhanced accuracy, factual consistency, and clinical relevance in radiological applications. Significance of Hallucination Mitigation Our findings reinforce RAG's critical role in hallucination mitigation, a fundamental requirement for deploying AI systems in clinical workflows. The complete elimination of hallucinations (reducing incidence from 8–0%) in our RAG-enhanced model represents a notable achievement, especially considering previous studies reporting hallucination rates of 28.6–39.6% in medical-specialized models 21 . Incorrect recommendations, particularly in contraindication management and contrast dosing, pose substantial patient safety risks. The success of our approach likely stems from developing a robust knowledge base centered specifically on ICM consultation, enabling the model to generate responses with clear evidential support. This achievement is particularly significant for safety-critical applications where even occasional hallucinations could potentially endanger patients. Privacy and Legal Considerations Given the legal and ethical constraints associated with cloud-based LLMs, particularly regarding PHI transmission under HIPAA and GDPR frameworks, our results suggest that RAG-enhanced locally deployable models provide a viable alternative that balances safety, speed, and privacy 4,22 . By processing data entirely within institutional networks, locally deployed LLMs substantially reduce information leakage risks, addressing a primary concern in healthcare AI adoption. While locally deployed models have traditionally faced performance limitations due to size constraints, our RAG-enhanced Llama 3.2 11B demonstrates that clinically useful AI support can be achieved in privacy-sensitive healthcare environments without compromising on essential performance metrics. This addresses the fundamental privacy-performance dilemma highlighted in our introduction, providing practical evidence for healthcare institutions evaluating AI implementation strategies. LLM-as-a-Judge: Efficacy and Limitations The observed divergence between human and LLM-based evaluators highlights an important consideration for future benchmarking frameworks. While LLM-based evaluators consistently ranked the RAG-enhanced Llama 3.2 11B higher than GPT-4o mini, the radiologist favored the latter. This may reflect the nuanced clinical judgment applied by human experts, including the handling of ambiguous cases and prioritization of patient safety 23,24 . Our analysis reported mean ranks, standard deviations, and 95% confidence intervals to complement the ordinal ranking data. Although such metrics derive from non-interval data, they are widely used in recent LLM benchmarking studies (e.g., MT-Bench, Chatbot Arena) as descriptive evaluator preference and ranking consistency indicators. These metrics provide a practical and interpretable summary of comparative model performance when interpreted with non-parametric tests and complete rank distributions. This divergence between evaluation methodologies underscores the need for hybrid evaluation strategies that combine quantitative and qualitative metrics, especially for high-stakes applications. Research on emergency medicine summary generation has shown that LLM-generated content rated highly by automated evaluations may contain subtle clinical utility and safety deficiencies only identifiable through expert clinician review 25 . Our hybrid approach offers valuable insights for refining methodologies used to evaluate LLMs in clinical contexts. Limitations and Future Directions Several limitations merit consideration. First, the study relied on simulated scenarios rather than real-world patient data, which may not capture the full complexity of clinical environments. Second, a single radiologist performed the expert evaluation, which may limit generalizability. Third, long-term maintenance of domain-specific knowledge bases and adaptation to evolving guidelines require further research. Future work should address the challenges of maintaining and updating RAG knowledge bases to incorporate evolving clinical guidelines, expanding evaluation to include multiple expert reviewers, and testing in real clinical workflows with appropriate privacy safeguards. Additionally, exploring RAG optimization for diverse radiological tasks beyond ICM consultation could further establish the utility of this approach in broader healthcare contexts. Conclusion In conclusion, this study provides evidence that RAG-enhanced local LLMs can serve as effective, privacy-preserving decision support tools in radiology. By achieving zero hallucination rates and response times faster than cloud-based alternatives, our approach demonstrates practicality for clinical implementation. As healthcare systems seek to adopt AI solutions that meet performance and compliance standards, retrieval-augmented locally deployable models offer a promising pathway for secure, scalable, and clinically meaningful deployment in safety-critical radiological applications. Declarations Author Contribution A.W. conceived the study, obtained funding, supervised all research activities, and drafted the initial manuscript; Y.T. curated and pre-processed consultation transcripts, performed statistical analyses, and co-wrote the Methods section; M.N. designed and maintained the AI environment, implemented core algorithms, and co-supervised technical execution; A.Ya. developed the knowledge-base architecture, advised on the RAG workflow, and reviewed technical accuracy; T.A. coordinated access to institutional resources, verified scenario relevance, and edited the Results section; A.H., Y.H., and J.K. jointly designed and validated the iodinated-contrast-media consultation scenarios, curated the chatbot knowledge base, prepared Supplementary Table 1 and key figures, and critically revised the manuscript; K.Sh. wrote statistical scripts, generated visualisations, and contributed to the Statistical Analysis subsection; K.Sa. verified experimental reproducibility, assisted model evaluation, and reviewed the manuscript; K.K. provided neuroradiological expertise, interpreted exemplar cases, and refined the clinical-implications paragraph; A.N. conducted the literature review on contrast-media guidelines, refined the consultation rubric, and contributed to manuscript editing; S.A. offered overall clinical guidance, critically revised all drafts, and gave final approval for submission. Data Availability All data generated and analysed during this study—including the 100 simulated iodinated-contrast-media consultation scenarios, model outputs, human rankings, and LLM-based rubric scores—are provided in the Supplementary Information files. Additional materials are available from the corresponding author upon reasonable request. References Liu Z, Zhong A, Li Y, others. Radiology-GPT: A Large Language Model for Radiology. arXiv. 2023; doi: 10.48550/arxiv.2306.08666. Nakaura T, Ito R, Ueda D, others. The impact of large language models on radiology: a guide for radiologists on the latest innovations in AI. Jpn J Radiol. 2024;1–12. doi: 10.1007/s11604-024-01552-0. Wada A, Akashi T, Shih G, others. Optimizing GPT-4 Turbo Diagnostic Accuracy in Neuroradiology through Prompt Engineering and Confidence Thresholds. Diagnostics. 2024;14(14):1541. doi: 10.3390/diagnostics14141541. Abbasi N, Smith DA. Cybersecurity in Healthcare: Securing Patient Health Information (PHI). J Knowl Learn Sci Technol ISSN: 2959-6386 (online). 2024;3(3):278–287. doi: 10.60087/jklst.vol3.n3.p.278-287. Mohan A, Ye M, Franke H, others. Securing AI Inference in the Cloud: Is CPU-GPU Confidential Computing Ready? 2024 IEEE 17th Int Conf Cloud Comput (CLOUD). 2024. p. 164–175. doi: 10.1109/cloud62652.2024.00028. Bedi S, Liu Y, Orr-Ewing L, others. A Systematic Review of Testing and Evaluation of Healthcare Applications of Large Language Models (LLMs). medRxiv. 2024;2024.04.15.24305869. doi: 10.1101/2024.04.15.24305869. Lewis P, Perez E, Piktus A, others. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. arXiv. 2020; doi: 10.48550/arxiv.2005.11401. Telenti A, Auli M, Hie BL, others. Large language models for science and medicine. Eur J Clin Investig. 2024;54(6):e14183. doi: 10.1111/eci.14183. Davenport MS, Chu P, Szczykutowicz TP, Smith-Bindman R. Comparison of Strategies to Conserve Iodinated Intravascular Contrast Media for Computed Tomography During a Shortage. JAMA. 2022;328(5):476–478. doi: 10.1001/jama.2022.9879. CADTH. Optimizing the Use of Iodinated Contrast Media for CT: Managing Shortages and Planning for a Sustainable and Secure Supply. Can J Heal Technol. 2023;3(3). doi: 10.51731/cjht.2023.613. Sessa M, Rossi C, Rafaniello C, et al. Campania preventability assessment committee: a focus on the preventability of the contrast media adverse drug reactions. Expert Opin Drug Saf. 2016;15(sup2):51–59. doi: 10.1080/14740338.2016.1226280. Siewert B, Brook OR, Hochman M, Eisenberg RL. Impact of Communication Errors in Radiology on Patient Care, Customer Satisfaction, and Work-Flow Efficiency. Am J Roentgenol. 2016;206(3):573–579. doi: 10.2214/ajr.15.15117. Dify.AI · The Innovation Engine for Generative AI Applications. 2023. https://dify.ai/ . Beckett KR, Moriarity AK, Langer JM. Safe Use of Contrast Media: What the Radiologist Needs to Know. Radiographics. 2015;35(6):1738–1750. doi: 10.1148/rg.2015150033. Isaka Y, Hayashi H, Aonuma K, et al. Guideline on the use of iodinated contrast media in patients with kidney disease 2018. Clin Exp Nephrol. 2020;24(1):1–44. doi: 10.1007/s10157-019-01750-5. Thomsen HS, Morcos SK, ESUR. ESUR guidelines on contrast media. Abdom Imaging. 2006;31(2):131–140. doi: 10.1007/s00261-005-0380-y. Liu Y, Iter D, Xu Y, Wang S, Xu R, Zhu C. G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment. arXiv. 2023; doi: 10.48550/arxiv.2303.16634. Gupta S, Ranjan R, Singh SN. A Comprehensive Survey of Retrieval-Augmented Generation. arXiv. 2024; doi: 10.48550/arxiv.2410.12837. Mortaheb M, Khojastepour MAA, Chakradhar ST, Ulukus S. RAG-Check: Evaluating Multimodal Retrieval-Augmented Generation. arXiv. 2025; doi: 10.48550/arxiv.2501.03995. Nguyen Q, Nguyen D-A, Dang K, et al. Advancing Question-Answering in Ophthalmology with Retrieval-Augmented Generation (RAG): Benchmarking Open-source and Proprietary Large Language Models. medRxiv. 2024;2024.11.18.24317510. doi: 10.1101/2024.11.18.24317510. Chelli M, Descamps J, Lavoué V, et al. Hallucination Rates and Reference Accuracy of ChatGPT and Bard for Systematic Reviews: Comparative Analysis. J Méd Internet Res. 2024;26:e53164. doi: 10.2196/53164. Ng MY, Helzer J, Pfeffer MA, Seto T, Hernandez-Boussard T. Development of secure infrastructure for advancing generative artificial intelligence research in healthcare at an academic medical center. J Am Méd Inform Assoc. 2025;32(3):586–588. doi: 10.1093/jamia/ocaf005. Iglesia ID la, Goenaga I, Ramirez-Romero J, Villa-Gonzalez JM, Goikoetxea J, Barrena A. Ranking Over Scoring: Towards Reliable and Robust Automated Evaluation of LLM-Generated Medical Explanatory Arguments. arXiv. 2024; doi: 10.48550/arxiv.2409.20565. Gu J, Jiang X, Shi Z, et al. A Survey on LLM-as-a-Judge. arXiv. 2024; doi: 10.48550/arxiv.2411.15594. Szymanski A, Ziems N, Eicher-Miller HA, Li TJ-J, Jiang M, Metoyer RA. Limitations of the LLM-as-a-Judge Approach for Evaluating LLM Outputs in Expert Knowledge Tasks. arXiv. 2024; doi: 10.48550/arxiv.2410.20266. Tables Table 1. Technical Specifications of Evaluated Language Models for ICM Consultation and Response Evaluation. LLM Name Use Case Deployment Developer Model Scale Description Availability GPT-4o mini ICM Consultation Cloud OpenAI Large (MoE) Optimized for speed 2024-07 Gemini 2.0 Flash ICM Consultation Cloud Google Large Fast-response model 2024-12 Claude 3.5 Haiku ICM Consultation Cloud Anthropic Medium–Large Lightweight inference 2024-10 Llama 3.2 11B ICM Consultation Local Meta 11B Base on-premise model 2024-09 GPT-4o Evaluator Cloud OpenAI Very Large Full-scale evaluator 2024-05 Gemini 2.0 Flash Exp. Evaluator Cloud Google Very Large Reasoning- tuned version 2024-12 Claude 3.5 Sonnet Evaluator Cloud Anthropic Very Large Mid-tier evaluator 2024-06 Note: Deployment — Cloud = vendor-hosted API only; Local = weights are deployable on-premise, though in this study Meta Llama 3.2 11B was run on GroqCloud (TLS-encrypted, no-retention) through the same Dify pipeline for parity. Model scale — “Large” ≈ tens-to-hundreds B parameters, “Very Large” the upper end; Llama’s 11 B count is official. “ MoE” = Mixture-of-Experts architecture. Table 2. Sample Knowledge Entry from the RAG Knowledge Base. Topic Question-Answer Format Precautions for Contrast Media Administration in Patients with Renal Dysfunction Q: What are the important precautions and considerations when administering contrast media to patients with renal dysfunction? A: • Pre-existing renal dysfunction is considered a risk factor for acute renal failure caused by iodinated contrast media, with eGFR now commonly used as the preferred indicator of renal function • Patients with eGFR below 30mL/min/1.73m² face significantly increased risk and require sufficient preventive measures • Contrast-induced nephropathy (CIN) is defined as an increase in serum creatinine level by 0.5mg/dL or 25% or more from baseline within 72 hours after contrast media administration • Key risk factors for CIN include advanced age, diabetes with chronic kidney disease (CKD), use of diuretics (especially loop diuretics), and use of NSAIDs • Appropriate hydration protocols using normal saline solution or sodium bicarbonate solution are recommended as preventive measures • Repeated contrast-enhanced CT scans within a short period (24-48 hours) should be avoided Example of a structured entry used in the retrieval-augmented generation (RAG) system, formatted as a clinical question and answer. This entry addresses precautions for administering contrast media to patients with renal dysfunction. Table 3. Model Performance Rankings by Radiologist and LLM Evaluators. Rank distribution (%) of five LLMs across 100 ICM consultation scenarios as rated by one radiologist and three LLM-based evaluators (GPT-4o, Gemini 2.0 Flash Experimental, and Claude 3.5 Sonnet). Color intensity indicates frequency of rankings (dark blue: ≥60%, medium blue: 40-59%, light blue: 30-39%, pale blue: 10-29%, white: <10%). Mean ranks with standard deviations and 95% confidence intervals are shown on the right. Lower ranks indicate better performance. Table 4. Clinical Errors and Corresponding RAG Corrections in Contrast Media Use. Category Clinical Scenario Initial Output Error RAG Correction Dosage Adult contrast dosage 1.5–3.0 mL/kg, max 300mg iodine/kg Adjusted to 1.5–2.0 mL/kg, 300 mgI/mL Dosage 3-year-old child (15kg) 5–10 mL total dosage 22.5–30 mL (weight-based) Contraindications Iodinated contrast risk Incorrectly mentioned NSF risk Identified contrast-induced nephropathy Contraindications Hyperthyroidism Omitted thyroid crisis risk Added crisis risk explanation Contraindications Pregnancy and CT Incomplete information Added fetal thyroid impact Contraindications β-blockers and anaphylaxis No management advice Added glucagon use info Incomplete Orders Metformin use unknown Mentioned lactic acidosis risk only Added stepwise protocol Avoidable Use Gallstone suspicion Recommended contrast CT Recommended ultrasound first Avoidable Use Emphysema, hemorrhage, fracture Correctly advised no contrast Maintained accuracy Summary of initial output errors made by the base LLM, corresponding corrections by the RAG-enhanced model, and whether the errors were fully addressed. Cases are categorized by clinical task type (e.g., dosage, contraindications, appropriateness). Additional Declarations No competing interests reported. Supplementary Files Supplementalmaterialquery100.csv Supplement.docx Cite Share Download PDF Status: Published Journal Publication published 02 Jul, 2025 Read the published version in npj Digital Medicine → Version 1 posted Editorial decision: Revision requested 22 May, 2025 Reviews received at journal 22 May, 2025 Reviewers agreed at journal 07 May, 2025 Reviews received at journal 04 May, 2025 Reviewers agreed at journal 01 May, 2025 Reviewers invited by journal 30 Apr, 2025 Editor assigned by journal 30 Apr, 2025 Submission checks completed at journal 29 Apr, 2025 First submitted to journal 27 Apr, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6538325","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":451691817,"identity":"e79961f2-5435-4508-8336-9dc110e3b165","order_by":0,"name":"Akihiko Wada","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6UlEQVRIiWNgGAWjYLACHgYbZgiLDUIZEKElDaiFmTQthxlQtOAF/A08hh/e1JxnN7jdf/BxQZkNA3/7AYbiAjxaJA7wGEvOOXab2eDOYWbjGefSGCTOJDAYz8Bnzf23G6R52IBabiSzSfO2AV14g4HBmAePDvkDvJt/8/w7B9LC/hukRZ6QFoMDvNuAhh8A28IM0mJASIvhAf5vlnP7kpklbyQbS/OcS+MxPJPYgNcvcgfYkm+8+WaXzHcj8eFnnjIbObnjh48Z4wsxGEiGMYBOYmwzJkIHgx0yh/kxMVpGwSgYBaNgxAAAP59FXD3rR68AAAAASUVORK5CYII=","orcid":"","institution":"Juntendo University","correspondingAuthor":true,"prefix":"","firstName":"Akihiko","middleName":"","lastName":"Wada","suffix":""},{"id":451691818,"identity":"d76fe00e-0d02-427b-8437-7f8699b976f4","order_by":1,"name":"Yuya Tanaka","email":"","orcid":"","institution":"The University of Tokyo","correspondingAuthor":false,"prefix":"","firstName":"Yuya","middleName":"","lastName":"Tanaka","suffix":""},{"id":451691819,"identity":"0757bb18-dd81-488f-80a9-37393370410d","order_by":2,"name":"Mitsuo Nishizawa","email":"","orcid":"","institution":"Juntendo University Urayasu Hospital","correspondingAuthor":false,"prefix":"","firstName":"Mitsuo","middleName":"","lastName":"Nishizawa","suffix":""},{"id":451691820,"identity":"b5e0282f-0ec8-4596-aea6-575d8a7a54b1","order_by":3,"name":"Akira Yamamoto","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Akira","middleName":"","lastName":"Yamamoto","suffix":""},{"id":451691821,"identity":"665a8777-1866-4ecb-90f5-e1288f451727","order_by":4,"name":"Toshiaki Akashi","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Toshiaki","middleName":"","lastName":"Akashi","suffix":""},{"id":451691823,"identity":"cebd645b-fbcd-42a6-ab7b-96970644a0c1","order_by":5,"name":"Akifumi Hagiwara","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Akifumi","middleName":"","lastName":"Hagiwara","suffix":""},{"id":451691825,"identity":"a2a3eeb5-bc8c-42d6-b5c7-308ce891fb96","order_by":6,"name":"Yayoi Hayakawa","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Yayoi","middleName":"","lastName":"Hayakawa","suffix":""},{"id":451691828,"identity":"6b071925-29f4-4ea5-89e0-20cca53cef27","order_by":7,"name":"Junko Kikuta","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Junko","middleName":"","lastName":"Kikuta","suffix":""},{"id":451691829,"identity":"85fc9c9a-2309-4baa-80c4-8da811d8879b","order_by":8,"name":"Keigo Shimoji","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Keigo","middleName":"","lastName":"Shimoji","suffix":""},{"id":451691830,"identity":"8f949b4a-50f9-4d8d-a1f6-e4e3e4e28088","order_by":9,"name":"Katsuhiro Sano","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Katsuhiro","middleName":"","lastName":"Sano","suffix":""},{"id":451691831,"identity":"e894ab48-205f-4693-acfd-5e05354b5cc0","order_by":10,"name":"Koji Kamagata","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Koji","middleName":"","lastName":"Kamagata","suffix":""},{"id":451691832,"identity":"7aec099d-038b-43d1-ab42-65dfcc77ffdf","order_by":11,"name":"Atsushi Nakanishi","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Atsushi","middleName":"","lastName":"Nakanishi","suffix":""},{"id":451691834,"identity":"82d3dd95-86d3-4efe-8ee6-b2bfe3f30580","order_by":12,"name":"Shigeki Aoki","email":"","orcid":"","institution":"Juntendo University","correspondingAuthor":false,"prefix":"","firstName":"Shigeki","middleName":"","lastName":"Aoki","suffix":""}],"badges":[],"createdAt":"2025-04-27 06:08:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6538325/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6538325/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41746-025-01802-z","type":"published","date":"2025-07-02T15:58:04+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":82276547,"identity":"02792021-e372-4a85-8212-fab9e1e5062e","added_by":"auto","created_at":"2025-05-08 14:46:33","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":53487,"visible":true,"origin":"","legend":"\u003ch4\u003eEvaluation Workflow for Language Model Responses.\u003c/h4\u003e\n\u003cp\u003eSchematic of the evaluation pipeline. Each clinical query was simultaneously processed by five LLMs (three cloud-based and two locally deployable models). The generated responses were anonymized and evaluated by a radiologist and three LLM-based evaluators using both human ranking and rubric-based scoring.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6538325/v1/4c8971c278330fdecb1ecb1e.png"},{"id":82278244,"identity":"4c60f4ee-d2b0-4719-b9ab-561b47f44d16","added_by":"auto","created_at":"2025-05-08 14:54:33","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":45034,"visible":true,"origin":"","legend":"\u003ch4\u003eRetrieval-Augmented Generation (RAG) Pipeline for Local LLMs.\u003c/h4\u003e\n\u003cp\u003eDiagram illustrating the RAG implementation. Domain-specific knowledge is embedded and indexed using semantic and keyword-based retrieval. Retrieved context chunks are ranked and integrated into the prompt before inference by the local LLM.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6538325/v1/4dac2ef6930134752b113bec.png"},{"id":82276554,"identity":"8ec6e4a6-75e4-470a-90b8-ba97574c0a54","added_by":"auto","created_at":"2025-05-08 14:46:33","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":59537,"visible":true,"origin":"","legend":"\u003ch4\u003eModel Rankings by Human and Automated Evaluators.\u003c/h4\u003e\n\u003cp\u003eThe mean ranks of five LLMs across 100 ICM consultation scenarios were evaluated by a radiologist and three LLM-based scorers. Lower values indicate better performance. Error bars represent 95% confidence intervals.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6538325/v1/b798b1854ad3f3bc378fcfff.png"},{"id":82276562,"identity":"24cff9a0-209e-4e5b-a5a1-71e677ec8ff1","added_by":"auto","created_at":"2025-05-08 14:46:33","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":107466,"visible":true,"origin":"","legend":"\u003ch4\u003eMultidimensional Performance Metrics Across LLMs.\u003c/h4\u003e\n\u003cp\u003eRadar charts comparing five LLMs across six key evaluation domains: clinical accuracy, safety, response structure, professional communication, practical applicability, and response time.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6538325/v1/91af589ae1c9d945deeab23f.png"},{"id":82278243,"identity":"65b3eaf9-b431-4122-ab34-8169493b3a35","added_by":"auto","created_at":"2025-05-08 14:54:33","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":38176,"visible":true,"origin":"","legend":"\u003ch4\u003eComparative Response Times of Evaluated LLMs.\u003c/h4\u003e\n\u003cp\u003eBoxplots showing response time distributions for five LLMs. Each point represents an individual measurement. Though with greater variability, the RAG-enhanced Llama model showed faster average responses than cloud-based models.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6538325/v1/64bfdcb2af531f96cfc169a0.png"},{"id":86179116,"identity":"29bc40b7-7a6c-4a79-93ed-367369833e84","added_by":"auto","created_at":"2025-07-07 16:15:54","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1360241,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6538325/v1/cb03bac7-7b13-42c0-bada-45d25ad4b667.pdf"},{"id":82276551,"identity":"544a3e85-434e-4d2b-8c17-835812107c56","added_by":"auto","created_at":"2025-05-08 14:46:33","extension":"csv","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":21366,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementalmaterialquery100.csv","url":"https://assets-eu.researchsquare.com/files/rs-6538325/v1/91578583a4d27860a764a2c4.csv"},{"id":82276555,"identity":"d5da711d-17ca-4133-b563-0c503da1c471","added_by":"auto","created_at":"2025-05-08 14:46:33","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":25158,"visible":true,"origin":"","legend":"","description":"","filename":"Supplement.docx","url":"https://assets-eu.researchsquare.com/files/rs-6538325/v1/aa28fc0c82276ef31b042323.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Retrieval-Augmented Generation Elevates Local LLM Quality in Radiology Contrast-Media Consultation","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLarge language models (LLMs) have rapidly expanded across domains, demonstrating significant potential to enhance human capabilities and workflow efficiency in specialized fields that require domain-specific knowledge. In medicine, LLMs are promising to summarize medical literature, support clinical reasoning, and extract key information from medical records \u003csup\u003e1,2\u003c/sup\u003e. In radiology, recent studies highlight LLMs' potential to improve report quality, workflow efficiency, and diagnostic accuracy \u003csup\u003e3,4\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDespite this potential, healthcare implementation of LLMs faces unique challenges centered on a fundamental dilemma: balancing patient data confidentiality with advanced AI capabilities. Cloud-based LLMs offer superior reasoning abilities through large-scale training data but require transmitting sensitive medical information to external servers, raising substantial concerns under regulations like HIPAA and GDPR \u003csup\u003e5\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eConversely, locally deployed LLMs ensure data privacy but typically demonstrate inferior performance due to constraints in model size and computational resources \u003csup\u003e6\u003c/sup\u003e. This performance gap is particularly pronounced in specialized medical domains where domain-specific knowledge and nuanced understanding of clinical guidelines are essential.\u003c/p\u003e \u003cp\u003eRetrieval-augmented generation (RAG) has emerged as a promising approach to address this trade-off. By integrating external domain-specific knowledge into the response generation process, RAG enables LLMs to leverage specialized information beyond their training data, reduce hallucinations through authoritative source grounding, and dynamically update knowledge without additional model fine-tuning \u003csup\u003e7\u003c/sup\u003e. RAG combines high accuracy with robust data privacy when implemented within a locally deployed environment, potentially offering a solution to the privacy-performance dilemma in healthcare AI.\u003c/p\u003e \u003cp\u003eWhile RAG's effectiveness has been demonstrated in general knowledge retrieval tasks, its utility in highly specialized medical fields, particularly radiology, remains underexplored. Previous studies have identified knowledge base quality and retrieval efficiency as key determinants of RAG performance in medical applications \u003csup\u003e8,9\u003c/sup\u003e, but few have evaluated its application in time-sensitive clinical decision support scenarios \u003csup\u003e10\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTo systematically assess locally deployed RAG-enhanced LLMs in a specialized medical context, this study focuses on iodinated contrast media (ICM) consultations—a crucial radiological task requiring specialized knowledge and real-time decision-making. CT examinations have become fundamental to modern diagnostic imaging, with annual scan volumes in the United States increasing from approximately 3\u0026nbsp;million in 1980 to 91\u0026nbsp;million in 2019 \u003csup\u003e11\u003c/sup\u003e. Notably, 40–48% of all CT scans involve ICM to enhance visualization of vascular and organ structures, making appropriate contrast administration essential to radiological practice.\u003c/p\u003e \u003cp\u003eRecent challenges have amplified the importance of precise ICM consultation. The COVID-19 pandemic caused severe global contrast agent shortages, necessitating careful assessment of contrast-enhanced scan appropriateness \u003csup\u003e12\u003c/sup\u003e. Additionally, ICM-related adverse events range from mild allergic reactions to severe nephrotoxicity, making risk stratification crucial for patient safety. Research indicates that 7.5% of ICM-related complications could be prevented through appropriate consultation \u003csup\u003e13\u003c/sup\u003e, while communication errors in contrast media ordering account for 13.9% of radiology workflow inefficiencies \u003csup\u003e14\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eEffective ICM consultation requires specialized radiological knowledge, clinical experience, and dynamic risk assessment capabilities. LLMs could potentially assist radiologists by automating risk assessment and protocol selection, thereby reducing workload and enhancing decision-making. The integration of facility-specific protocols and recently updated guidelines through RAG could be particularly valuable, as contrast media administration often involves institution-specific rules that cannot be captured in general model training.\u003c/p\u003e \u003cp\u003eThis study aims to evaluate the performance of RAG-enhanced lightweight local LLMs in ICM consultation compared to leading cloud-based models. We developed and assessed a RAG-implemented locally deployable model (Llama 3.2 11B) against three cloud-based LLMs (GPT-4o mini, Gemini 2.0 Flash, Claude 3.5 Haiku) using 100 simulated clinical scenarios requiring ICM risk assessment and protocol selection. Both radiologists and automated LLM evaluators evaluated performance through a comprehensive assessment, examining clinical accuracy, safety, response structure, communication quality, practical applicability, and response time.\u003c/p\u003e \u003cp\u003eBy quantifying performance differences between cloud-based and locally deployed models and assessing whether RAG can substantially enhance locally deployable model performance while maintaining data privacy, this research provides practical evidence to guide healthcare institutions in selecting and implementing AI systems for clinical decision support.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e1. Consultation Scenarios Development\u003c/p\u003e\u003cp\u003eWe designed 100 simulated consultation scenarios covering a diverse range of iodinated contrast media (ICM) use cases, including appropriateness, contraindications, incomplete orders, and avoidable contrast use. These scenarios systematically reflected five key categories in radiological practice: (1) appropriateness consultations for specific clinical indications, (2) optimal contrast agent selection and protocols, (3) contraindication identification and risk assessment, (4) recognition of incomplete ordering information, and (5) identification of cases where contrast could be avoided.\u003c/p\u003e\u003cp\u003eScenarios were initially generated using GPT-4o and Claude 3.5 Sonnet with structured prompts for clinical authenticity. To ensure realism and guideline compliance, two radiologists (4 and 30 years of experience, respectively) independently reviewed and revised each scenario according to current ACR and ESUR guidelines. Special attention was given to incorporating realistic clinical nuances and decision-making challenges commonly encountered in ICM consultations. The complete set of scenarios is available in the supplemental materials.\u003c/p\u003e\u003cp\u003e2. Language Models Configuration\u003c/p\u003e\u003cp\u003eWe benchmarked five LLMs—three cloud-based models (GPT-4o mini from OpenAI, Gemini 2.0 Flash from Google, and Claude 3.5 Haiku from Anthropic) accessed through their public REST APIs, and two locally deployable models (Llama 3.2 11B from Meta, as a baseline and RAG-enhanced variant). Llama 3.2 11B is designed as a locally deployable, lightweight parameter model ideal for on-premises computing environments. An example of the local execution environment for the Llama 3.2 11B model is as follows: NVIDIA RTX 3090 GPU (24 GB VRAM; Graphics Processing Unit), Intel Core i7-12700K CPU (8 cores; Central Processing Unit), and 32 GB RAM (Random Access Memory). This study deployed the Llama 3.2 11B via GroqCloud's hosted inference API to facilitate direct performance comparisons with the cloud-native models.\u003c/p\u003e\u003cp\u003eAll models were deployed via the Dify platform (version 0.10.0) with standardized parameters, including consistent temperature settings (0.2) and single-shot response generation \u003csup\u003e15\u003c/sup\u003e. For each model, we recorded both the responses and the response times while maintaining all other default settings for consistent comparison.\u003c/p\u003e\u003cp\u003eFor the automated evaluation process, we employed three high-performance models (GPT-4o, Gemini 2.0 Flash Thinking Experimental, and Claude 3.5 Sonnet) as evaluators. Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e1\u003c/span\u003e documents the technical specifications for all models used in this study, including parameter sizes, deployment characteristics, and implementation details.\u003c/p\u003e\u003c/div\u003e\u003cp\u003e3. Retrieval-Augmented Generation Implementation\u003c/p\u003e\u003cp\u003eFor the RAG-enhanced model, we compiled a knowledge base from authoritative sources, including the ACR Manual on Contrast Media, ESUR guidelines, institutional protocols, and relevant literature \u003csup\u003e16–18\u003c/sup\u003e. Knowledge entries were initially generated in a question-answer format using Claude 3.5 Sonnet and were subsequently reviewed, modified, and organized by two radiologists (Table\u0026nbsp;2).\u003c/p\u003e\u003cp\u003eThe knowledge base was structured as segments with a maximum of 600 tokens, comprising 66 chunks with an average of 337 characters per segment. These segments were transformed into high-dimensional embeddings using OpenAI's text-embedding-3-large model and indexed in \"High Quality\" mode. The system employed a hybrid search method that combined semantic vector search with keyword-based retrieval.\u003c/p\u003e\u003cp\u003eFor each clinical query, the system retrieved four relevant context fragments (TopK = 4) based on cosine similarity calculations. These retrieved fragments were then ranked according to their relevance scores and incorporated into the prompt structure for the Llama 3.2 11B model (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e4. Evaluation Methodology\u003c/p\u003e\u003cp\u003eWe used a two-tier evaluation strategy: (i) human expert review by a board-certified radiologist, and (ii) automated scoring by three large LLM “judges.”\u003c/p\u003e\u003cp\u003eFor the human tier, the radiologist reviewed all responses in a blinded manner, ranking the anonymized responses from 1st (best) to 5th (worst) based on clinical appropriateness and practical applicability, without considering response time. In addition to rubric-based scoring, each response was screened for hallucinations, defined as clinically incorrect or guideline-inconsistent statements that could alter patient management (e.g., wrong dosage, missing contraindication). Hallucination presence (yes/no) was recorded per scenario, and the frequencies were compared between models using χ² (with Yates correction) and Fisher’s exact tests.\u003c/p\u003e\u003cp\u003eFor automated LLM “judge” evaluation, we utilized three high-performance LLM evaluators within a G-E VAL-like framework \u003csup\u003e19\u003c/sup\u003e. Each response was systematically compared against reference standards developed by radiology experts, using six predefined criteria:\u003c/p\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eClinical Accuracy: Alignment with established medical guidelines and factual correctness\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eSafety: Proper identification of contraindications and emphasis on patient safety measures\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eResponse Structure: Clarity and logical organization of the answer\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eProfessional Communication: Appropriate use of medical terminology and professional language\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003ePractical Applicability: Provision of actionable clinical recommendations\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eResponse Time: Performance against predefined speed thresholds\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e\u003cp\u003eLLM evaluators assessed the responses based on these criteria, with final rankings determined by the total combined score. In cases of tied scores, models were assigned the same higher rank. Comprehensive details regarding the evaluation prompts and scoring rubrics are available in the Supplementary Materials.\u003c/p\u003e\u003cp\u003e5. Statistical Analysis\u003c/p\u003e\u003cp\u003eAll analyses were conducted in Python 3.9.13 with scipy 1.10.0 and statsmodels 0.13.5. Because the Shapiro–Wilk test revealed significant departures from normality (p \u0026lt; 0.05) in the rank distributions, we exclusively applied non-parametric procedures. Overall differences among the five LLMs were examined with the Friedman test, which accounts for the repeated-measures nature of the 100 shared consultation scenarios. Pairwise contrasts were then assessed with the Nemenyi post-hoc test; for the two comparisons of primary interest—GPT-4o mini versus the RAG-enhanced Llama, and baseline versus RAG Llama—we also report Mann-Whitney U (Wilcoxon rank-sum) statistics with Bonferroni-adjusted significance thresholds. Effect sizes are expressed as the difference in mean ranks (Δ), accompanied by Wilcoxon Z values and two-sided p values. Descriptively, we present mean rank ± standard deviation and 95% confidence intervals to visualise performance spread and inter-evaluator agreement, following conventions established in recent LLM benchmarks such as MT-Bench and Chatbot Arena.\u003c/p\u003e\u003cp\u003e6. Ethical Considerations\u003c/p\u003e\u003cp\u003eThis study did not require Institutional Review Board approval as it utilized simulated clinical scenarios rather than actual patient data. All consultations were fictional cases designed to reflect common clinical situations without incorporating any protected health information.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e1. Overall Model Performance and Rankings\u003c/p\u003e \u003cp\u003eA global Friedman test across the five LLMs confirmed significant differences in median ranks for the 100 iodinated-contrast-media scenarios (χ\u0026sup2; = 78.4, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Post-hoc Nemenyi comparisons showed that the three cloud models clustered at the top: Gemini 2.0 Flash was dominant (radiologist mean rank\u0026thinsp;=\u0026thinsp;1.36; first in 74% of cases), followed by Claude 3.5 Haiku (2.09) and GPT-4o mini (3.21). The baseline Llama 3.2 11B was ranked last in 86% of cases (mean 4.80). RAG lifted its performance to a mean rank of 3.54, close to GPT-4o mini\u0026rsquo;s 3.21 (Δ = \u0026minus;\u0026thinsp;1.26; Z = \u0026minus;\u0026thinsp;7.26, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Automated LLM evaluators confirmed the overall cloud supremacy, again ranking Gemini 2.0 Flash and Claude 3.5 Haiku first and second, respectively. Within the remaining models, however, they placed the RAG-enhanced Llama 3.2 11B as 3rd, ahead of GPT-4o mini (Δ = \u0026minus;\u0026thinsp;0.44 to \u0026minus;\u0026thinsp;0.82; Z\u0026thinsp;=\u0026thinsp;+\u0026thinsp;2.56 to +\u0026thinsp;5.43; p\u0026thinsp;\u0026le;\u0026thinsp;0.009). This reversal relative to the radiologist\u0026rsquo;s preference illustrates a systematic divergence between human and LLM-based judgements.\u003c/p\u003e \u003cp\u003e2. Inter-Evaluator Agreement and Assessment Patterns\u003c/p\u003e \u003cp\u003eIn three LLM evaluators, Claude 3.5 Sonnet exhibited the highest agreement with the radiologist at 82.2%, suggesting its assessments most closely align with human expert judgment. GPT-4o\u0026rsquo;s evaluation also showed a relatively high agreement rate of 75.5% compared to the radiologist. Across all evaluators, there was broad consensus on the top performer (Gemini 2.0 Flash, 81.6% agreement) and bottom performer (base Llama 3.2 11B), while mid-tier models showed more variable assessments.\u003c/p\u003e \u003cp\u003eThe most significant evaluation divergence appeared in mid-tier rankings. All three LLM evaluators consistently ranked the RAG-enhanced Llama 3.2 11B higher than GPT-4o mini, contrary to the radiologist's assessment. This divergence was most pronounced with GPT-4o, which ranked the RAG-enhanced Llama second (mean rank: 2.53) and GPT-4o mini fourth (mean rank: 3.35).\u003c/p\u003e \u003cp\u003e3. Clinical Quality and Hallucination Mitigation\u003c/p\u003e \u003cp\u003eQualitative assessment by a radiologist revealed that hallucinations\u0026mdash;particularly in contrast dosing and contraindication recommendations\u0026mdash;were present in 8% of responses from the base Llama model and were eliminated (0%) in the RAG-enhanced version (χ\u0026sup2;₍Yates₎ = 6.38, p\u0026thinsp;=\u0026thinsp;0.012; Fisher p\u0026thinsp;=\u0026thinsp;0.0068). These hallucinations typically involved incorrect ICM protocols, including improper contraindication identification and dosage recommendations, which could potentially impact patient safety. None of the cloud-based models exhibited hallucinations in the evaluated scenarios.\u003c/p\u003e \u003cp\u003eAmong the 100 cases, 54% showed marked improvement with RAG\u0026mdash;most notably in safety-critical content such as contraindication management and precise dosing guidance\u0026mdash;while an additional 33% demonstrated minor gains involving non-critical factual details. These improvements were concentrated in cases that required specialized knowledge of institutional protocols or recently updated guidelines.\u003c/p\u003e \u003cp\u003e4. Multidimensional Performance Assessment\u003c/p\u003e \u003cp\u003eRadar chart analysis of LLM evaluations indicated substantial gains in clinical accuracy, safety, and applicability for the RAG-enhanced model, with modest changes in communication and structure (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). The most significant improvements occurred in domains directly related to patient safety and clinical decision-making.\u003c/p\u003e \u003cp\u003eSpecifically, clinical accuracy scores improved from a mean of 14.2 to 21.5 points (out of 25), and safety considerations improved from 11.8 to 18.7 points (out of 20). Professional communication showed moderate enhancement (13.4 to 16.8 out of 20), while response structure remained relatively stable (7.8 to 8.3 out of 10), as the base model already demonstrated adequate formatting capabilities.\u003c/p\u003e \u003cp\u003e5. Response Time Analysis\u003c/p\u003e \u003cp\u003eIn terms of response time, the RAG-enhanced model (2.58s) remained faster on average than all cloud-based models (GPT-4o mini: 4.87s, Claude 3.5 Haiku: 7.25s, Gemini 2.0 Flash: 3.14s), despite increased latency compared to its base version (1.31s) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). The difference was statistically significant for all comparisons, including Gemini 2.0 Flash (t = \u0026minus;\u0026thinsp;3.96, p\u0026thinsp;=\u0026thinsp;0.00011).\u003c/p\u003e \u003cp\u003eResponse time variability differed notably across models. Claude 3.5 Haiku demonstrated the most consistent performance with a coefficient of variation (CV) of 10.76%, while the RAG-enhanced Llama 3.2 11B showed the highest variability (CV\u0026thinsp;=\u0026thinsp;44.19%), likely due to fluctuations in knowledge retrieval latency. The base Llama 3.2 11B (CV\u0026thinsp;=\u0026thinsp;37.40%), GPT-4o mini (CV\u0026thinsp;=\u0026thinsp;31.21%), and Gemini 2.0 Flash (CV\u0026thinsp;=\u0026thinsp;26.43%) exhibited intermediate variability. While the RAG-enhanced model's response time was more variable, the trade-off was acceptable given the substantial performance improvements in clinical accuracy and safety.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eRAG Technology in Radiological Applications\u003c/p\u003e \u003cp\u003eThis study demonstrates that a retrieval-augmented local LLM can achieve clinically reliable performance in a safety-critical radiological use case while preserving patient privacy. By integrating curated guideline-based knowledge into a lightweight LLM, we observed significant improvements in clinical accuracy, risk stratification, and communication quality\u0026mdash;criteria that are central to iodinated contrast media (ICM) decision-making. Notably, the RAG-enhanced model eliminated hallucinations observed in the base model and achieved faster average response times than all evaluated cloud-based LLMs, suggesting feasibility for real-time use in clinical settings.\u003c/p\u003e \u003cp\u003eOur findings align with recent research demonstrating RAG's potential to enhance LLM performance in specialized medical domains. Gupta et al. (2024) emphasized RAG's application in medical knowledge retrieval, showing improvements in accuracy and reliability\u003csup\u003e20\u003c/sup\u003e. Mortaheb et al. (2025) introduced a multimodal evaluation framework ensuring contextual relevance \u003csup\u003e9\u003c/sup\u003e, while Nguyen et al. (2024) reported significant accuracy improvements in clinical question-answering when incorporating RAG into language models\u003csup\u003e8\u003c/sup\u003e. These studies collectively confirm that integrating domain-specific knowledge improves evidence-based reasoning capabilities of LLMs, leading to enhanced accuracy, factual consistency, and clinical relevance in radiological applications.\u003c/p\u003e \u003cp\u003eSignificance of Hallucination Mitigation\u003c/p\u003e \u003cp\u003eOur findings reinforce RAG's critical role in hallucination mitigation, a fundamental requirement for deploying AI systems in clinical workflows. The complete elimination of hallucinations (reducing incidence from 8\u0026ndash;0%) in our RAG-enhanced model represents a notable achievement, especially considering previous studies reporting hallucination rates of 28.6\u0026ndash;39.6% in medical-specialized models \u003csup\u003e21\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIncorrect recommendations, particularly in contraindication management and contrast dosing, pose substantial patient safety risks. The success of our approach likely stems from developing a robust knowledge base centered specifically on ICM consultation, enabling the model to generate responses with clear evidential support. This achievement is particularly significant for safety-critical applications where even occasional hallucinations could potentially endanger patients.\u003c/p\u003e \u003cp\u003ePrivacy and Legal Considerations\u003c/p\u003e \u003cp\u003eGiven the legal and ethical constraints associated with cloud-based LLMs, particularly regarding PHI transmission under HIPAA and GDPR frameworks, our results suggest that RAG-enhanced locally deployable models provide a viable alternative that balances safety, speed, and privacy \u003csup\u003e4,22\u003c/sup\u003e. By processing data entirely within institutional networks, locally deployed LLMs substantially reduce information leakage risks, addressing a primary concern in healthcare AI adoption.\u003c/p\u003e \u003cp\u003eWhile locally deployed models have traditionally faced performance limitations due to size constraints, our RAG-enhanced Llama 3.2 11B demonstrates that clinically useful AI support can be achieved in privacy-sensitive healthcare environments without compromising on essential performance metrics. This addresses the fundamental privacy-performance dilemma highlighted in our introduction, providing practical evidence for healthcare institutions evaluating AI implementation strategies.\u003c/p\u003e \u003cp\u003eLLM-as-a-Judge: Efficacy and Limitations\u003c/p\u003e \u003cp\u003eThe observed divergence between human and LLM-based evaluators highlights an important consideration for future benchmarking frameworks. While LLM-based evaluators consistently ranked the RAG-enhanced Llama 3.2 11B higher than GPT-4o mini, the radiologist favored the latter. This may reflect the nuanced clinical judgment applied by human experts, including the handling of ambiguous cases and prioritization of patient safety \u003csup\u003e23,24\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eOur analysis reported mean ranks, standard deviations, and 95% confidence intervals to complement the ordinal ranking data. Although such metrics derive from non-interval data, they are widely used in recent LLM benchmarking studies (e.g., MT-Bench, Chatbot Arena) as descriptive evaluator preference and ranking consistency indicators. These metrics provide a practical and interpretable summary of comparative model performance when interpreted with non-parametric tests and complete rank distributions.\u003c/p\u003e \u003cp\u003eThis divergence between evaluation methodologies underscores the need for hybrid evaluation strategies that combine quantitative and qualitative metrics, especially for high-stakes applications. Research on emergency medicine summary generation has shown that LLM-generated content rated highly by automated evaluations may contain subtle clinical utility and safety deficiencies only identifiable through expert clinician review \u003csup\u003e25\u003c/sup\u003e. Our hybrid approach offers valuable insights for refining methodologies used to evaluate LLMs in clinical contexts.\u003c/p\u003e \u003cp\u003eLimitations and Future Directions\u003c/p\u003e \u003cp\u003eSeveral limitations merit consideration. First, the study relied on simulated scenarios rather than real-world patient data, which may not capture the full complexity of clinical environments. Second, a single radiologist performed the expert evaluation, which may limit generalizability. Third, long-term maintenance of domain-specific knowledge bases and adaptation to evolving guidelines require further research.\u003c/p\u003e \u003cp\u003eFuture work should address the challenges of maintaining and updating RAG knowledge bases to incorporate evolving clinical guidelines, expanding evaluation to include multiple expert reviewers, and testing in real clinical workflows with appropriate privacy safeguards. Additionally, exploring RAG optimization for diverse radiological tasks beyond ICM consultation could further establish the utility of this approach in broader healthcare contexts.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, this study provides evidence that RAG-enhanced local LLMs can serve as effective, privacy-preserving decision support tools in radiology. By achieving zero hallucination rates and response times faster than cloud-based alternatives, our approach demonstrates practicality for clinical implementation. As healthcare systems seek to adopt AI solutions that meet performance and compliance standards, retrieval-augmented locally deployable models offer a promising pathway for secure, scalable, and clinically meaningful deployment in safety-critical radiological applications.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eA.W. conceived the study, obtained funding, supervised all research activities, and drafted the initial manuscript; Y.T. curated and pre-processed consultation transcripts, performed statistical analyses, and co-wrote the Methods section; M.N. designed and maintained the AI environment, implemented core algorithms, and co-supervised technical execution; A.Ya. developed the knowledge-base architecture, advised on the RAG workflow, and reviewed technical accuracy; T.A. coordinated access to institutional resources, verified scenario relevance, and edited the Results section; A.H., Y.H., and J.K. jointly designed and validated the iodinated-contrast-media consultation scenarios, curated the chatbot knowledge base, prepared Supplementary Table 1 and key figures, and critically revised the manuscript; K.Sh. wrote statistical scripts, generated visualisations, and contributed to the Statistical Analysis subsection; K.Sa. verified experimental reproducibility, assisted model evaluation, and reviewed the manuscript; K.K. provided neuroradiological expertise, interpreted exemplar cases, and refined the clinical-implications paragraph; A.N. conducted the literature review on contrast-media guidelines, refined the consultation rubric, and contributed to manuscript editing; S.A. offered overall clinical guidance, critically revised all drafts, and gave final approval for submission.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eAll data generated and analysed during this study\u0026mdash;including the 100 simulated iodinated-contrast-media consultation scenarios, model outputs, human rankings, and LLM-based rubric scores\u0026mdash;are provided in the Supplementary Information files. Additional materials are available from the corresponding author upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eLiu Z, Zhong A, Li Y, others. Radiology-GPT: A Large Language Model for Radiology. arXiv. 2023; doi: 10.48550/arxiv.2306.08666.\u003c/li\u003e\n\u003cli\u003eNakaura T, Ito R, Ueda D, others. The impact of large language models on radiology: a guide for radiologists on the latest innovations in AI. Jpn J Radiol. 2024;1\u0026ndash;12. doi: 10.1007/s11604-024-01552-0.\u003c/li\u003e\n\u003cli\u003eWada A, Akashi T, Shih G, others. Optimizing GPT-4 Turbo Diagnostic Accuracy in Neuroradiology through Prompt Engineering and Confidence Thresholds. Diagnostics. 2024;14(14):1541. doi: 10.3390/diagnostics14141541.\u003c/li\u003e\n\u003cli\u003eAbbasi N, Smith DA. Cybersecurity in Healthcare: Securing Patient Health Information (PHI). J Knowl Learn Sci Technol ISSN: 2959-6386 (online). 2024;3(3):278\u0026ndash;287. doi: 10.60087/jklst.vol3.n3.p.278-287.\u003c/li\u003e\n\u003cli\u003eMohan A, Ye M, Franke H, others. Securing AI Inference in the Cloud: Is CPU-GPU Confidential Computing Ready? 2024 IEEE 17th Int Conf Cloud Comput (CLOUD). 2024. p. 164\u0026ndash;175. doi: 10.1109/cloud62652.2024.00028.\u003c/li\u003e\n\u003cli\u003eBedi S, Liu Y, Orr-Ewing L, others. A Systematic Review of Testing and Evaluation of Healthcare Applications of Large Language Models (LLMs). medRxiv. 2024;2024.04.15.24305869. doi: 10.1101/2024.04.15.24305869.\u003c/li\u003e\n\u003cli\u003eLewis P, Perez E, Piktus A, others. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. arXiv. 2020; doi: 10.48550/arxiv.2005.11401.\u003c/li\u003e\n\u003cli\u003eTelenti A, Auli M, Hie BL, others. Large language models for science and medicine. Eur J Clin Investig. 2024;54(6):e14183. doi: 10.1111/eci.14183.\u003c/li\u003e\n\u003cli\u003eDavenport MS, Chu P, Szczykutowicz TP, Smith-Bindman R. Comparison of Strategies to Conserve Iodinated Intravascular Contrast Media for Computed Tomography During a Shortage. JAMA. 2022;328(5):476\u0026ndash;478. doi: 10.1001/jama.2022.9879.\u003c/li\u003e\n\u003cli\u003eCADTH. Optimizing the Use of Iodinated Contrast Media for CT: Managing Shortages and Planning for a Sustainable and Secure Supply. Can J Heal Technol. 2023;3(3). doi: 10.51731/cjht.2023.613.\u003c/li\u003e\n\u003cli\u003eSessa M, Rossi C, Rafaniello C, et al. Campania preventability assessment committee: a focus on the preventability of the contrast media adverse drug reactions. Expert Opin Drug Saf. 2016;15(sup2):51\u0026ndash;59. doi: 10.1080/14740338.2016.1226280.\u003c/li\u003e\n\u003cli\u003eSiewert B, Brook OR, Hochman M, Eisenberg RL. Impact of Communication Errors in Radiology on Patient Care, Customer Satisfaction, and Work-Flow Efficiency. Am J Roentgenol. 2016;206(3):573\u0026ndash;579. doi: 10.2214/ajr.15.15117.\u003c/li\u003e\n\u003cli\u003eDify.AI \u0026middot; The Innovation Engine for Generative AI Applications. 2023. \u003ca data-fr-linked=\"true\" href=\"https://dify.ai/\"\u003ehttps://dify.ai/\u003c/a\u003e.\u003c/li\u003e\n\u003cli\u003eBeckett KR, Moriarity AK, Langer JM. Safe Use of Contrast Media: What the Radiologist Needs to Know. Radiographics. 2015;35(6):1738\u0026ndash;1750. doi: 10.1148/rg.2015150033.\u003c/li\u003e\n\u003cli\u003eIsaka Y, Hayashi H, Aonuma K, et al. Guideline on the use of iodinated contrast media in patients with kidney disease 2018. Clin Exp Nephrol. 2020;24(1):1\u0026ndash;44. doi: 10.1007/s10157-019-01750-5.\u003c/li\u003e\n\u003cli\u003eThomsen HS, Morcos SK, ESUR. ESUR guidelines on contrast media. Abdom Imaging. 2006;31(2):131\u0026ndash;140. doi: 10.1007/s00261-005-0380-y.\u003c/li\u003e\n\u003cli\u003eLiu Y, Iter D, Xu Y, Wang S, Xu R, Zhu C. G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment. arXiv. 2023; doi: 10.48550/arxiv.2303.16634.\u003c/li\u003e\n\u003cli\u003eGupta S, Ranjan R, Singh SN. A Comprehensive Survey of Retrieval-Augmented Generation. arXiv. 2024; doi: 10.48550/arxiv.2410.12837.\u003c/li\u003e\n\u003cli\u003eMortaheb M, Khojastepour MAA, Chakradhar ST, Ulukus S. RAG-Check: Evaluating Multimodal Retrieval-Augmented Generation. arXiv. 2025; doi: 10.48550/arxiv.2501.03995.\u003c/li\u003e\n\u003cli\u003eNguyen Q, Nguyen D-A, Dang K, et al. Advancing Question-Answering in Ophthalmology with Retrieval-Augmented Generation (RAG): Benchmarking Open-source and Proprietary Large Language Models. medRxiv. 2024;2024.11.18.24317510. doi: 10.1101/2024.11.18.24317510.\u003c/li\u003e\n\u003cli\u003eChelli M, Descamps J, Lavou\u0026eacute; V, et al. Hallucination Rates and Reference Accuracy of ChatGPT and Bard for Systematic Reviews: Comparative Analysis. J Méd Internet Res. 2024;26:e53164. doi: 10.2196/53164.\u003c/li\u003e\n\u003cli\u003eNg MY, Helzer J, Pfeffer MA, Seto T, Hernandez-Boussard T. Development of secure infrastructure for advancing generative artificial intelligence research in healthcare at an academic medical center. J Am Méd Inform Assoc. 2025;32(3):586\u0026ndash;588. doi: 10.1093/jamia/ocaf005.\u003c/li\u003e\n\u003cli\u003eIglesia ID la, Goenaga I, Ramirez-Romero J, Villa-Gonzalez JM, Goikoetxea J, Barrena A. Ranking Over Scoring: Towards Reliable and Robust Automated Evaluation of LLM-Generated Medical Explanatory Arguments. arXiv. 2024; doi: 10.48550/arxiv.2409.20565.\u003c/li\u003e\n\u003cli\u003eGu J, Jiang X, Shi Z, et al. A Survey on LLM-as-a-Judge. arXiv. 2024; doi: 10.48550/arxiv.2411.15594.\u003c/li\u003e\n\u003cli\u003eSzymanski A, Ziems N, Eicher-Miller HA, Li TJ-J, Jiang M, Metoyer RA. Limitations of the LLM-as-a-Judge Approach for Evaluating LLM Outputs in Expert Knowledge Tasks. arXiv. 2024; doi: 10.48550/arxiv.2410.20266.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003ch4\u003e\u003cstrong\u003eTable 1.\u0026nbsp;\u003c/strong\u003eTechnical Specifications of Evaluated Language Models for ICM Consultation and Response Evaluation.\u003c/h4\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"589\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLLM Name\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 89px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eUse Case\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDeployment\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 73px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDeveloper\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel Scale\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDescription\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAvailability\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003eGPT-4o mini\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 89px;\"\u003e\n \u003cp\u003eICM Consultation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003eCloud\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 73px;\"\u003e\n \u003cp\u003eOpenAI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003eLarge (MoE)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003eOptimized for speed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2024-07\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003eGemini 2.0 Flash\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 89px;\"\u003e\n \u003cp\u003eICM Consultation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003eCloud\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 73px;\"\u003e\n \u003cp\u003eGoogle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003eLarge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003eFast-response model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2024-12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003eClaude 3.5 Haiku\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 89px;\"\u003e\n \u003cp\u003eICM Consultation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003eCloud\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 73px;\"\u003e\n \u003cp\u003eAnthropic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003eMedium\u0026ndash;Large\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003eLightweight inference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2024-10\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003eLlama 3.2 11B\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 89px;\"\u003e\n \u003cp\u003eICM Consultation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003eLocal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 73px;\"\u003e\n \u003cp\u003eMeta\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003e11B\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003eBase on-premise model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2024-09\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003eGPT-4o\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 89px;\"\u003e\n \u003cp\u003eEvaluator\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003eCloud\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 73px;\"\u003e\n \u003cp\u003eOpenAI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003eVery Large\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003eFull-scale evaluator\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2024-05\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003eGemini 2.0 Flash Exp.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 89px;\"\u003e\n \u003cp\u003eEvaluator\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003eCloud\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 73px;\"\u003e\n \u003cp\u003eGoogle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003eVery Large\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 93px;\"\u003e\n \u003cp\u003eReasoning- tuned version\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2024-12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003eClaude 3.5 Sonnet\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 89px;\"\u003e\n \u003cp\u003eEvaluator\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 88px;\"\u003e\n \u003cp\u003eCloud\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 73px;\"\u003e\n \u003cp\u003eAnthropic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 84px;\"\u003e\n \u003cp\u003eVery Large\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003eMid-tier evaluator\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2024-06\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eNote:\u003c/strong\u003e \u003cem\u003eDeployment \u0026mdash; \u003cstrong\u003eCloud\u003c/strong\u003e = vendor-hosted API only; \u003cstrong\u003eLocal\u003c/strong\u003e = weights are deployable on-premise, though in this study Meta Llama 3.2 11B was run on GroqCloud (TLS-encrypted, no-retention) through the same Dify pipeline for parity. Model scale \u0026mdash; \u0026ldquo;Large\u0026rdquo; \u0026asymp; tens-to-hundreds B parameters, \u0026ldquo;Very Large\u0026rdquo; the upper end; Llama\u0026rsquo;s 11 B count is official. \u0026ldquo;\u003c/em\u003e\u003cem\u003eMoE\u0026rdquo; = Mixture-of-Experts architecture.\u003c/em\u003e\u003c/p\u003e\n\u003ch4\u003eTable 2. Sample Knowledge Entry from the RAG Knowledge Base.\u003c/h4\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"629\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 151px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTopic\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 478px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQuestion-Answer Format\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 151px;\"\u003e\n \u003cp\u003ePrecautions for Contrast Media Administration in Patients with Renal Dysfunction\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 478px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ: What are the important precautions and considerations when administering contrast media to patients with renal dysfunction?\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eA:\u003c/strong\u003e \u0026bull; Pre-existing renal dysfunction is considered a risk factor for acute renal failure caused by iodinated contrast media, with eGFR now commonly used as the preferred indicator of renal function\u003c/p\u003e\n \u003cp\u003e\u0026bull; Patients with eGFR below 30mL/min/1.73m\u0026sup2; face significantly increased risk and require sufficient preventive measures\u003c/p\u003e\n \u003cp\u003e\u0026bull; Contrast-induced nephropathy (CIN) is defined as an increase in serum creatinine level by 0.5mg/dL or 25% or more from baseline within 72 hours after contrast media administration\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026bull; Key risk factors for CIN include advanced age, diabetes with chronic kidney disease (CKD), use of diuretics (especially loop diuretics), and use of NSAIDs\u003c/p\u003e\n \u003cp\u003e\u0026bull; Appropriate hydration protocols using normal saline solution or sodium bicarbonate solution are recommended as preventive measures\u003c/p\u003e\n \u003cp\u003e\u0026bull; Repeated contrast-enhanced CT scans within a short period (24-48 hours) should be avoided\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eExample of a structured entry used in the retrieval-augmented generation (RAG) system, formatted as a clinical question and answer. This entry addresses precautions for administering contrast media to patients with renal dysfunction.\u003c/p\u003e\n\u003ch4\u003eTable 3. Model Performance Rankings by Radiologist and LLM Evaluators.\u003c/h4\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"677\" height=\"463\"\u003e\u003c/p\u003e\n\u003cp\u003eRank distribution (%) of five LLMs across 100 ICM consultation scenarios as rated by one radiologist and three LLM-based evaluators (GPT-4o, Gemini 2.0 Flash Experimental, and Claude 3.5 Sonnet). Color intensity indicates frequency of rankings (dark blue: \u0026ge;60%, medium blue: 40-59%, light blue: 30-39%, pale blue: 10-29%, white: \u0026lt;10%). Mean ranks with standard deviations and 95% confidence intervals are shown on the right. Lower ranks indicate better performance.\u003c/p\u003e\n\u003ch4\u003eTable 4. Clinical Errors and Corresponding RAG Corrections in Contrast Media Use.\u003c/h4\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"586\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCategory\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eClinical Scenario\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInitial Output Error\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRAG Correction\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDosage\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003eAdult contrast dosage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003e1.5\u0026ndash;3.0 mL/kg, max 300mg iodine/kg\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eAdjusted to 1.5\u0026ndash;2.0 mL/kg, 300 mgI/mL\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDosage\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003e3-year-old child (15kg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003e5\u0026ndash;10 mL total dosage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e22.5\u0026ndash;30 mL (weight-based)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eContraindications\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003eIodinated contrast risk\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003eIncorrectly mentioned NSF risk\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eIdentified contrast-induced nephropathy\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eContraindications\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003eHyperthyroidism\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003eOmitted thyroid crisis risk\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eAdded crisis risk explanation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eContraindications\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003ePregnancy and CT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003eIncomplete information\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eAdded fetal thyroid impact\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eContraindications\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003e\u0026beta;-blockers and anaphylaxis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003eNo management advice\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eAdded glucagon use info\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIncomplete Orders\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003eMetformin use unknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003eMentioned lactic acidosis risk only\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eAdded stepwise protocol\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAvoidable Use\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003eGallstone suspicion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003eRecommended contrast CT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eRecommended ultrasound first\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAvoidable Use\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 165px;\"\u003e\n \u003cp\u003eEmphysema, hemorrhage, fracture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\n \u003cp\u003eCorrectly advised no contrast\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eMaintained accuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eSummary of initial output errors made by the base LLM, corresponding corrections by the RAG-enhanced model, and whether the errors were fully addressed. Cases are categorized by clinical task type (e.g., dosage, contraindications, appropriateness).\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6538325/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6538325/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eLarge language models (LLMs) excel at text reasoning, but hospital use is curtailed by privacy rules and limited medical tuning. We asked whether retrieval-augmented generation (RAG) can lift a lightweight, locally deployable LLM to near-cloud performance. In 100 simulated iodinated-contrast-media consultations we compared Llama 3.2-11B (baseline and RAG) with three cloud-based models\u0026mdash;GPT-4o mini, Gemini 2.0 Flash and Claude 3.5 Haiku. A blinded radiologist ranked the five replies per case, and three LLM judges scored accuracy, safety, structure, tone, applicability and latency. RAG eliminated hallucinations (0% vs 8%; χ\u0026sup2;₍Yates₎ = 6.38, p\u0026thinsp;=\u0026thinsp;0.012) and improved mean rank by 1.3 (Z = \u0026minus;\u0026thinsp;4.82, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), narrowing most of the quality gap while remaining faster (2.6 s vs 4.9\u0026ndash;7.3 s). The LLM judges, who weighted latency, preferred the RAG-enhanced Llama, whereas the radiologist ranked GPT-4o mini third. RAG thus enables near-cloud performance on-premise without exposing patient data.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e","manuscriptTitle":"Retrieval-Augmented Generation Elevates Local LLM Quality in Radiology Contrast-Media Consultation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-08 14:46:28","doi":"10.21203/rs.3.rs-6538325/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-05-22T23:45:31+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-22T16:46:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"151799892460582236345107594168949072007","date":"2025-05-07T22:56:28+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-04T15:51:25+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"129509803683072107355975055526755636451","date":"2025-05-01T07:04:09+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-05-01T02:38:39+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-05-01T00:27:37+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-30T03:50:37+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Digital Medicine","date":"2025-04-27T05:58:05+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5598e4e3-9b81-49aa-ab58-57270170a4e2","owner":[],"postedDate":"May 8th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":48044572,"name":"Health sciences/Health care/Medical imaging"},{"id":48044573,"name":"Health sciences/Health care"}],"tags":[],"updatedAt":"2025-07-07T16:05:18+00:00","versionOfRecord":{"articleIdentity":"rs-6538325","link":"https://doi.org/10.1038/s41746-025-01802-z","journal":{"identity":"npj-digital-medicine","isVorOnly":false,"title":"npj Digital Medicine"},"publishedOn":"2025-07-02 15:58:04","publishedOnDateReadable":"July 2nd, 2025"},"versionCreatedAt":"2025-05-08 14:46:28","video":"","vorDoi":"10.1038/s41746-025-01802-z","vorDoiUrl":"https://doi.org/10.1038/s41746-025-01802-z","workflowStages":[]},"version":"v1","identity":"rs-6538325","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6538325","identity":"rs-6538325","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}