In the paper, the novel method for generating personalized magnetic resonance imaging (MRI) reports based on key phrases is introduced. In order to demonstrate the method, an application with multiple large language model (LLM) agents was written. For establishing connection between different agents, graph-based software architecture was used. The graph consists of the following nodes. (1) The data retriever node. In this node relevant reports along with the metadata are obtained from the vector database. (2) The findings generation node. This node includes an implementation of two different variations of an MRI report’s findings section generation method. The first is based on merging already existed findings parts of radiology reports that were retrieved by key phrases into the new one. The second is based on reconstructing findings section so that it contains specific paragraphs retrieved by key phrases from the vector store. (3) The impression generation node. This node implements summarization of a findings section into an impression. The performance of a findings part generation was evaluated with BLEU, ROUGE and BERTScore metrics. Experiments have shown promising results in generation of findings parts: for the first approach mean ROUGE-L score was about 0.5, BERTScore was around 0.8, for the second approach mean ROUGE-L score was approximately 0.6, BERTScore was roughly 0.9. The first approach was around three times slower than the second. During the experiments, the dataset of 908 reports gathered from seven radiologists was used. Qualitative analysis was performed with five-point Likert scale questionary and statistically analyzed by means of Mann–Whitney U test.

1. [1] L. Luo, “Combining knowledge graph and artificial intelligence to conduct financial report quality detection research,” J. Adv. Comput. Intell. Intell. Inform., Vol.29, No.4, pp. 787-795, 2025. https://doi.org/10.20965/jaciii.2025.p0787
2. [2] M. Fareed, M. Fatima, J. Uddin, A. Ahmed, and M. A. Sattar, “A systematic review of ethical considerations of large language models in healthcare and medicine,” Frontiers in Digital Health, Vol.7, Article No.1653631, 2025. https://doi.org/10.3389/fdgth.2025.1653631
3. [3] T. Nakaura et al., “The impact of large language models on radiology: A guide for radiologists on the latest innovations in AI,” Japanese J. of Radiology, Vol.42, No.7, pp. 685-696, 2024. https://doi.org/10.1007/s11604-024-01552-0
4. [4] M. P. Hartung, I. C. Bickle, F. Gaillard, and J. P. Kanne, “How to create a great radiology report,” RadioGraphics, Vol.40, No.6, pp. 1658-1670, 2020. https://doi.org/10.1148/rg.2020200020
5. [5] Z. Sun et al., “Evaluating GPT-4 on impressions generation in radiology reports,” Radiology, Vol.307, No.5, Article No.e231259, 2023. https://doi.org/10.1148/radiol.231259
6. [6] A. Serapio et al., “An open-source fine-tuned large language model for radiological impression generation: A multi-reader performance study,” BMC Medical Imaging, Vol.24, Article No.254, 2024. https://doi.org/10.1186/s12880-024-01435-w
7. [7] L. Zhang et al., “Constructing a large language model to generate impressions from findings in radiology reports,” Radiology, Vol.312, No.3, Article No.e240885, 2024. https://doi.org/10.1148/radiol.240885
8. [8] Ș.-V. Voinea et al., “GPT-driven radiology report generation with fine-tuned Llama 3,” Bioengineering, Vol.11, No.10, Article No.1043, 2024. https://doi.org/10.3390/bioengineering11101043
9. [9] F. Zeng, Z. Lyu, Q. Li, and X. Li, “Enhancing LLMs for impression generation in radiology reports through a multi-agent system,” arXiv:2412.06828, 2024. https://doi.org/10.48550/arXiv.2412.06828
10. [10] P. Lewis et al., “Retrieval-augmented generation for knowledge-intensive NLP tasks,” Proc. of the 34th Int. Conf. on Neural Information Processing Systems, pp. 9459-9474, 2020.
11. [11] M. L. Saini, V. M. Shrimal, A. Garg, D. C. Sati, and S. P. K. Mygapula, “Medical diagnosis with RAG-LLMs: A hybrid approach for AI-driven healthcare,” IEEE 4th Int. Conf. for Advancement in Technology, 2025. https://doi.org/10.1109/ICONAT66879.2025.11362545
12. [12] H. M. T. Alam, D. Srivastav, M. A. Kadir, and D. Sonntag, “Towards interpretable radiology report generation via concept bottlenecks using a multi-agentic RAG,” Proc. of 47th European Conf. on Information Retrieval, Part 3, pp. 201-209, 2025. https://doi.org/10.1007/978-3-031-88714-7_18
13. [13] Ç. U. Öğdü, K. Arslanoğlu, and M. Karaköse, “An adaptive multi-agent LLM-based clinical decision support system integrating biomedical RAG and web intelligence,” IEEE Access, Vol.13, pp. 167390-167404, 2025. https://doi.org/10.1109/ACCESS.2025.3613340.
14. [14] J. Burrows, “‘Delta’: A measure of stylistic difference and a guide to likely authorship,” Literary and Linguistic Computing, Vol.17, No.3, pp. 267-287, 2002. https://doi.org/10.1093/llc/17.3.267
15. [15] FRIDA. https://huggingface.co/ai-forever/FRIDA [Accessed November 17, 2025]
16. [16] Chroma: Open-source search infrastructure for AI. https://www.trychroma.com/ [Accessed November 17, 2025]
17. [17] A. Yang et al., “Qwen3 technical report,” arXiv:2505.09388, 2025. https://doi.org/10.48550/arXiv.2505.09388
18. [18] A. Grattafiori et al., “The Llama 3 herd of models,” arXiv:2407.21783, 2024. https://doi.org/10.48550/arXiv.2407.21783
19. [19] C.-Y. Lin, “ROUGE: A package for automatic evaluation of summaries,” Text Summarization Branches Out (Proc. of the ACL-04 Workshop), pp. 74-81, 2004.
20. [20] K. Papineni, S. Roukos, T. Ward, and W.-J. Zhu, “Bleu: A method for automatic evaluation of machine translation,” Proc. of the 40th Annual Meeting of the Association for Computational Linguistics, pp. 311-318, 2002. https://doi.org/10.3115/1073083.1073135
21. [21] Y. Wu et al., “Google’s neural machine translation system: Bridging the gap between human and machine translation,” arXiv:1609.08144, 2016. https://doi.org/10.48550/arXiv.1609.08144
22. [22] T. Zhang, V. Kishore, F. Wu, K. Q. Weinberger, and Y. Artzi, “BERTScore: Evaluating text generation with BERT,” arXiv:1904.09675, 2019. https://doi.org/10.48550/arXiv.1904.09675
23. [23] J. A. Omiye, H. Gui, S. J. Rezaei, J. Zou, and R. Daneshjou, “Large language models in medicine: The potentials and pitfalls: A narrative review,” Annals of Internal Medicine, Vol.177, No.2, pp. 210-220, 2024. https://doi.org/10.7326/M23-2772
24. [24] S. A. Kostrov and M. P. Potapov, “Large language models in medicine: Current ethical challenges,” Medical Ethics, Vol.2025, No.2, pp. 22-31, 2025. https://doi.org/10.24075/medet.2025.008
25. [25] M. Cascella et al., “The breakthrough of large language models release for medical applications: 1-year timeline and perspectives,” J. of Medical Systems, Vol.48, Article No.22, 2024. https://doi.org/10.1007/s10916-024-02045-3
26. [26] S. M. C. A. Negri, “Robot as legal person: Electronic personhood in robotics and artificial intelligence,” Frontiers in Robotics and AI, Vol.8, Article No.789327, 2021. https://doi.org/10.3389/frobt.2021.789327
27. [27] S. Schmidgall et al., “Evaluation and mitigation of cognitive biases in medical language models,” npj Digital Medicine, Vol.7, Article No.295, 2024. https://doi.org/10.1038/s41746-024-01283-6
28. [28] N. Yadav et al., “Data privacy in healthcare: In the era of artificial intelligence,” Indian Dermatology Online J., Vol.14, No.6, pp. 788-792, 2023. https://doi.org/10.4103/idoj.idoj_543_23
29. [29] I. C. Wiest et al., “Anonymizing medical documents with local, privacy preserving large language models: The LLM-anonymizer,” medRxiv, 2024. https://doi.org/10.1101/2024.06.11.24308355