Figure - available from: European Journal of Nuclear Medicine and Molecular Imaging
This content is subject to copyright. Terms and conditions apply.
Workflow of the Chatbot System for Querying PET Imaging Reading Reports. The overall workflow of the proof-of-concept system designed for efficient querying of reading reports from a substantial dataset is illustrated. The system integrates the Retrieval-Augmented Generation (RAG) model with advanced language model technologies, natural language processing, and information retrieval techniques. The workflow demonstrates the process from user query input through to the delivery of the relevant reading report, showcasing the operational framework and interaction with different sources of reading reports
Source publication
Purpose
The potential of Large Language Models (LLMs) in enhancing a variety of natural language tasks in clinical fields includes medical imaging reporting. This pilot study examines the efficacy of a retrieval-augmented generation (RAG) LLM system considering zero-shot learning capability of LLMs, integrated with a comprehensive database of PET r...
Similar publications
Personas are hypothetical representations of real-world people used as storytelling tools to help designers identify the goals, constraints, and scenarios of particular user groups. A well-constructed persona can provide enough detail to trigger recognition and empathy while leaving room for varying interpretations of users. While a traditional per...
ChatGPT, or Chat Generative Pre-trained Transformer, developed by OpenAI, is a versatile chatbot known for generating human-like text responses. Since its launch in November 2022, it has sparked interest and debate. This bibliometric study aimed to explore ChatGPT-related publications using the Web of Science database from 2023 to June 2024. Origin...
This study examines the effectiveness of the ChatMu application in improving access to information for members of Muhammadiyah, a prominent socio-religious organization. The research employs a mixed-methods approach, combining qualitative and quantitative analyses to evaluate the application’s performance, usability, and user satisfaction. Findings...
Direct Preference Optimization (DPO) demonstrates the advantage of aligning a large language model with human preference using only an offline dataset. However, DPO has the limitation that the KL penalty, which prevents excessive deviation from the reference model, is static throughout the training process. Several methods try to turn this static K...
When speakers are exposed to a particular syntactic construction, it can prime them to use similar structures in their own speech. Previous studies have demonstrated that this syntactic priming effect is influenced by a multitude of linguistic, cognitive, and social factors. The current study evaluated the impact of perceived AI interlocutors on sy...
Citations
... An additional potential use of LLMs is improving references to prior reports, narrowing down differential diagnoses and supporting clinician decision making. The combination of a retrieval-augmented generation (RAG) LLM system with an extensive database of previous PET imaging reports from patients with breast cancer, lung cancer, and lymphoma enabled the identification of similar cases and the extraction of potential diagnoses based on those cases [31]. ...
Recently, there has been tremendous interest on the use of large language models (LLMs) in radiology. LLMs have been employed for various applications in cancer imaging, including improving reporting speed and accuracy via generation of standardized reports, automating the classification and staging of abnormal findings in reports, incorporating appropriate guidelines, and calculating individualized risk scores. Another use of LLMs is their ability to improve patient comprehension of imaging reports with simplification of the medical terms and possible translations to multiple languages. Additional future applications of LLMs include multidisciplinary tumor board standardizations, aiding patient management, and preventing and predicting adverse events (contrast allergies, MRI contraindications) and cancer imaging research. However, limitations such as hallucinations and variable performances could present obstacles to widespread clinical implementation. Herein, we present a review of the current and future applications of LLMs in cancer imaging, as well as pitfalls and limitations.
... The copyright holder for this preprint this version posted April 1, 2025. ; environments (AlGhadban et al., 2023;Alonso et al., 2024;Choi et al., 2024;. Specific applications were identified in ophthalmology diagnostics and the analysis of multimodal patient data, suggesting broader opportunities for integrating RAG AI into medical training and diagnostic processes (Upadhyaya et al., 2024). ...
Background: Retrieval-augmented generation (RAG) is an emerging artificial intelligence (AI) strategy that integrates encoded model knowledge with external data sources to enhance accuracy, transparency, and reliability. Unlike traditional large language models (LLMs), which are limited by static training data and potential misinformation, RAG dynamically retrieves and integrates relevant medical literature, clinical guidelines, and real-time data. Given the rapid adoption of AI in healthcare, this scoping review aims to systematically map the current applications, implementation challenges, and research gaps related to RAG in health professions.
Methods: A scoping review was conducted following the Joanna Briggs Institute (JBI) framework and reported using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Scoping Reviews (PRISMA-ScR) guidelines. A systematic search strategy was designed in collaboration with faculty and a research and education librarian to include PubMed, Scopus, Embase, Google Scholar, and Trip, covering studies published between January 2020 and August 2024. Eligible studies examined the use of RAG in healthcare. Studies were screened in two stages: title/abstract review followed by full-text assessment. Data extraction focused on study characteristics, applications of RAG, ethical and technical challenges, and proposed improvements.
Results: A total of 31 studies met inclusion criteria, with 90.32% published in 2024. Authors came from 17 countries with the most frequent publications coming from the USA (n = 15), China (n = 3), and the Republic of Korea (n = 3). Key applications included clinical decision support, healthcare education, and pharmacovigilance. Ethical concerns centered on data privacy, algorithmic bias, explainability, and potential overreliance on AI-generated recommendations. Bias mitigation strategies included dataset diversification, fine-tuning techniques, and expert oversight. Transparency measures such as structured citations, traceable information retrieval, and explainable diagnostic pathways were explored to enhance clinician trust in AI-generated outputs. Identified challenges included optimizing retrieval mechanisms, improving real-time integration, and standardizing validation frameworks.
Conclusion: RAG AI has the potential to improve clinical decision-making and healthcare education by addressing key limitations of traditional LLMs. However, significant challenges remain regarding ethical implementation, model reliability, and regulatory oversight. Future research should prioritize refining retrieval accuracy, strengthening bias mitigation strategies, and establishing standardized evaluation metrics. Responsible deployment of RAG-based systems requires interdisciplinary collaboration between AI researchers, clinicians, and policymakers to ensure ethical, transparent, and effective integration into healthcare workflows.
