Elamur Research⌕
☰
home / papers / 2026 / RadTimeline: Timeline Summarization for Longitudinal Radiological Lung Findings

RadTimeline: Timeline Summarization for Longitudinal Radiological Lung Findings

Summary

Radiologists and clinicians must follow how lung findings change across a patient's successive imaging reports. Doing this by hand is slow and mentally taxing, and free-text AI summaries are hard to verify. The authors recast multi-report summarization as building a two-dimensional timeline table. Columns are dated exams, cells hold individual lung findings from each report, and rows track one finding's evolution under a descriptive row header.

Their generation pipeline uses off-the-shelf LLMs without supervised training, in three stages:

  1. Extract findings from each report.
  2. Propose group names from the longitudinal record.
  3. Assign each finding to a group.

To evaluate this, they built RadTimeline, a small hand-corrected benchmark drawn from chest-related imaging reports. They also proposed metrics for finding factuality, overgeneration, name quality and grouping.

Experiments compare LLMs of different sizes and several prompting strategies. The intermediate naming step proves essential for good grouping. The strongest setup recovers nearly all relevant findings, though it adds some irrelevant ones, and groups findings about as well as human annotators. The format makes AI summaries easier to fact-check in high-stakes clinical settings.

Key points

  • RadTimeline dataset: 10 patients from a UW Medicine lung cancer surveillance cohort, with 65 chest-related reports (chest CT, chest X-ray, abdomen CT).
    • Each patient has 4–12 reports, with longitudinal text of 646–4625 tokens (mean 2055).
    • Six patients developed lung cancer within 3 years; four had no cancer record.
  • Annotation: gold timelines were made by hand-correcting LLM drafts, with two medical students and one medical doctor as annotators. Findings outside the lungs (e.g., the aorta) and device mentions are excluded.
  • Models and evaluation: models compared include GPT-4o and Llama 3.1 8B Instruct. Findings are scored with ROUGE-L plus overgeneration, alongside group name quality and grouping metrics.
  • Smaller model behavior: Llama 3.1 tends to produce more findings and group names, including content unrelated to the lungs.
  • Role of group names: generated group names matter most for weaker LLMs and for embedding-based grouping.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Topics: Summarization

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/32uhj8rx3u8f

@inproceedings{zhou2026radtimeline,
  title = {{RadTimeline: Timeline Summarization for Longitudinal Radiological Lung Findings}},
  author = {Zhou, Sitong and Yetisgen, Meliha and Ostendorf, Mari},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/32uhj8rx3u8f},
  url = {https://doi.org/10.63317/32uhj8rx3u8f}
}

References

25 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Shany Barhom, Vered Shwartz, Alon Eirew, Michael Bugert, Nils Reimers, and Ido Dagan. 2019. Revisiting joint modeling of cross-document entity and event coreference resolution. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4179–4189, Florence, Ital… find our card
  2. Avi Caciularu, Arman Cohan, Iz Beltagy, Matthew Peters, Arie Cattan, and Ido Dagan. 2021. CDLM: Cross-document language modeling. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 2648–2662, Punta Cana, Dominican Republic. Association for Computational Linguistics. find
  3. A. Chien, H. Tang, B. Jagessar, K. W. Chang, N. Peng, K. Nael, and N. Salamon. 2024. AI-assisted summarization of radiologic reports: Evaluating GPT3davinci, BARTcnn, LongT5booksum, LEDbooksum, LEDlegal, and LEDclinical. AJNR American Journal of Neuroradiology, 45(2):244–248. find
  4. Surabhi Datta, Hio Cheng Lam, Atieh Pajouhi, Sunitha Mogalla, and Kirk Roberts. 2022. A cross-document coreference dataset for longitudinal tracking across radiology reports. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 3686–3695, Marseille, France. European L… find
  5. Jean-Benoit Delbrouck, Pierre Chambon, Christian Bluethgen, Emily Tsai, Omar Almusa, and Curtis Langlotz. 2022. Improving the factual correctness of radiology report generation with semantic rewards. In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 4348–4360, Abu Dhabi… find
  6. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783. find
  7. Chen Huang and Guoxiu He. 2024. Text clustering as classification with LLMs. arXiv preprint, arXiv:2410.00927. — [arXiv:2410.00927](https://arxiv.org/abs/2410.00927) find
  8. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, et al. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276. find our card
  9. Sameer Tajdin Khanna, Adam Dejl, Kibo Yoon, Steven QH Truong, Hanh Duong, Agustina Saenz, and Pranav Rajpurkar. 2023. RadGraph2: Modeling disease progression in radiology reports via hierarchical information extraction. In MLHC, pages 381–402. find
  10. K. Lee, N. J. Dobbins, B. McInnes, M. Yetisgen, and Ö. Uzuner. 2021. Transferability of neural network clinical deidentification systems. J Am Med Inform Assoc. find
  11. Tejas Sudharshan Mathai, Boah Kim, Oana M. Stroie, and Ronald M. Summers. 2025. Privacy–preserving large language model for matching findings and tracking interval changes in longitudinal radiology reports. Journal of Imaging Informatics in Medicine. U.S. Government work; foreign copyright may apply… find
  12. Qingkai Min, Qipeng Guo, Xiangkun Hu, Songfang Huang, Zheng Zhang, and Yue Zhang. 2024. Synergetic event understanding: A collaborative approach to cross-document event coreference resolution with large language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational L… find our card
  13. Sewon Min, Kalpesh Krishna, Xinxi Lyu, Mike Lewis, Wen-tau Yih, Pang-Wei Koh, Mohit Iyyer, Luke Zettlemoyer, and Hannaneh Hajishirzi. 2023. FActScore: Fine-grained atomic evaluation of factual precision in long form text generation. In Proceedings of the 2023 Conference on Empirical Methods in Natur… find our card
  14. Niklas Muennighoff, Nouamane Tazi, Loic Magne, and Nils Reimers. 2023. MTEB: Massive text embedding benchmark. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 2014–2037, Dubrovnik, Croatia. Association for Computational Linguistic… find our card
  15. Namu Park, Kevin Lybarger, Giridhar Kaushik Ramachandran, Spencer Lewis, Aashka Damani, Özlem Uzuner, Martin Gunn, and Meliha Yetisgen. 2024. A novel corpus of annotated medical imaging reports and information extraction results using BERT-based language models. In Proceedings of the 2024 Joint Inte… find
  16. Sameer Pradhan, Alessandro Moschitti, Nianwen Xue, Olga Uryupina, and Yuchen Zhang. 2012. CoNLL-2012 shared task: Modeling multilingual unrestricted coreference in OntoNotes. In Joint Conference on EMNLP and CoNLL – Shared Task, pages 1–40, Jeju Island, Korea. Association for Computational Linguisti… find our card
  17. Kawshik Manikantan Sundar, Shubham Toshniwal, Makarand Tapaswi, and Vineet Gandhi. 2024. Major entity identification: A generalizable alternative to coreference resolution. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 11679– 11695, Miami, Florida,… find our card
  18. Dave Van Veen, Cara Van Uden, Maayane Attias, Anuj Pareek, Christian Bluethgen, Malgorzata Polacin, Wah Chiu, Jean-Benoit Delbrouck, Juan Zambrano Chaves, Curtis Langlotz, Akshay Chaudhari, and John Pauly. 2023. RadAdapt: Radiology report summarization via lightweight domain adaptation of large lang… find
  19. Dave Van Veen, Cara Van Uden, Louis Blankemeier, Jean-Benoit Delbrouck, Asad Aali, Christian Bluethgen, Anuj Pareek, Malgorzata Polacin, Eduardo Pontes Reis, Anna Seehofnerová, Nidhi Rohatgi, Poonam Hosamani, William Collins, Neera Ahuja, Curtis P. Langlotz, Jason Hom, Sergios Gatidis, John Pauly, a… find our card
  20. R. Verma, E. Alsentzer, Z. Strasser, L. Chang, K. Roman, E. Gershanik, C. Hernandez, M. Linares, J. Rodriguez, D. Thakral, O. Unlu, J. You, L. Zhou, and D. Bates. 2025. Verifiable summarization of electronic health records using large language models to support chart review. medRxiv. Preprint. find
  21. Liang Wang, Nan Yang, Xiaolong Huang, Linjun Yang, Rangan Majumder, and Furu Wei. 2024. Improving text embeddings with large language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 11897– 11916, Bangkok, Thailand. Ass… find our card
  22. Joonho Yang, Seunghyun Yoon, ByeongJeong Kim, and Hwanhee Lee. 2024. FIZZ: Factual inconsistency detection by zoom-in summary and zoom-out document. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 30–45, Miami, Florida, USA. Association for Computatio… find
  23. Fang Zeng, Zhiliang Lyu, Quanzheng Li, and Xiang Li. 2024. Enhancing LLMs for impression generation in radiology reports through a multi-agent system. arXiv preprint, arXiv:2412.06828. — [arXiv:2412.06828](https://arxiv.org/abs/2412.06828) find
  24. Sihang Zeng, Yujuan Fu, Sitong Zhou, Zixuan Yu, Lucas Jing Liu, Jun Wen, Matthew Thompson, Ruth Etzioni, and Meliha Yetisgen. 2025. Traj-CoA: Patient trajectory modeling via chain-of-agents for lung cancer risk prediction. In The Second Workshop on GenAI for Health: Potential, Trust, and Policy Comp… find
  25. Jin Zhao, Nianwen Xue, and Bonan Min. 2023. Cross-document event coreference resolution: Instruct humans or instruct GPT? In Proceedings of the 27th Conference on Computational Natural Language Learning (CoNLL), pages 561– 574, Singapore. Association for Computational Linguistics. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhou-2026-radtimeline-timeline-summarization-longitudinal-radiological/