Enhancing Rare Disease Diagnosis from Clinical Notes through Chain-of-Thought and Retrieval-Augmented Generation
DOI:
https://doi.org/10.66238/0jwgdr90Keywords:
Rare Diseases, Clinical Diagnostics, Natural Language Processing, Medical Informatics, Retrieval-Augmented GenerationAbstract
The diagnosis of rare diseases remains one of the most formidable challenges in modern medical practice, often characterized by a prolonged diagnostic odyssey that imposes significant physical, psychological, and financial burdens on patients. Electronic health records contain vast amounts of unstructured clinical narratives that harbor critical clues for identifying these elusive conditions. However, the sheer volume of data and the cognitive limits of human practitioners often obscure these signals. Recent advancements in large language models offer unprecedented capabilities in natural language understanding, yet their direct application in medical diagnostics is hindered by issues such as knowledge hallucination and a lack of grounding in authoritative medical literature. This paper proposes a novel framework that integrates retrieval-augmented generation with chain-of-thought reasoning to enhance the accuracy and interpretability of rare disease diagnosis from unstructured clinical notes. By dynamically retrieving pertinent medical literature from curated knowledge bases and explicitly modeling the step-by-step deductive reasoning process characteristic of expert physicians, the proposed architecture significantly mitigates hallucinations while improving diagnostic precision. Our comprehensive analysis demonstrates that this dual-component approach outperforms traditional machine learning baselines and standalone large language models across multiple evaluation metrics. Furthermore, the explicit reasoning pathways generated by the system provide crucial transparency, fostering clinical trust and offering a viable pathway toward the integration of advanced artificial intelligence diagnostic support systems in real-world healthcare settings.
References
1. Bulten, W., Pinckaers, H., van Boven, H., Vink, R., de Bel, T., van Ginneken, B., et al. (2020). Artificial intelligence for diagnosis and Gleason grading of prostate cancer: The PANDA challenge. Nature Medicine, 26(9), 1544–1551.
2. Xie, C., Zhu, D., Wang, Z., Zhang, H., & Wei, Z. (2026). Explainable AI for Mental Health Detection from Social Media: A Comparative Study of Traditional Machine Learning and a Large Language Model. Available at SSRN 6429778.
3. Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542(7639), 115–118.
4. Dong, J., Koniusz, P., Zhang, Y., Zhu, H., Liu, W., Qu, X., & Ong, Y. S. (2025, October). Improving zero-shot adversarial robustness in vision-language models by closed-form alignment of adversarial path simplices. In Forty-second International Conference on Machine Learning.
5. Lv, Qi, et al. "F1: A vision-language-action model bridging understanding and generation to actions." arXiv preprint arXiv:2509.06951 (2025).
6. Dong, J., Wang, Y., Lai, J., & Xie, X. (2023). Restricted black-box adversarial attack against deepfake face swapping. IEEE Transactions on Information Forensics and Security, 18, 2596-2608.
7. Ren, Y., Wu, D., Tong, Y., López-DeFede, A., & Gareau, S. (2023). Issue of data imbalance on low birthweight baby outcomes prediction and associated risk factors identification: establishment of benchmarking key machine learning models with data rebalancing strategies. Journal of medical Internet research, 25, e44081.
8. Xie, C., Zhu, D., Wang, Z., Zhang, H., & Wei, Z. (2026). Compliance-Aware Discharge Agent for Auditable ICU Discharge Planning: A Pilot Feasibility Study Using Structured eICU Records. Available at SSRN 6429758.
9. Guo, Hanzhong, et al. "Real-time identity defenses against malicious personalization of diffusion models." arXiv preprint arXiv:2412.09844 (2024).
10. Lei, C., Tao, X., & Xiao, Y. (2023, May). Dietary Iodine intake and related factors among secondary school students in Macao. In Healthcare (Vol. 11, No. 10, p. 1472). MDPI.
11. Wang, Y., Ma, X., Huang, X., Li, S., Yu, X., Huang, S., ... & Yin, S. (2026). Integrating environmental factors and genetic variants in machine learning to assess occupational noise impact on health. Ecotoxicology and Environmental Safety, 311, 119805.
12. Zhu, Di, et al. "RaX-Crash: A Resource Efficient and Explainable Small Model Pipeline with an Application to City Scale Injury Severity Prediction." arXiv preprint arXiv:2512.07848 (2025).
13. Chu, Y. B., Li, J., Jia, P., Cui, J., Zhang, R., Kang, X., ... & Zhang, S. (2021). Irf1-and Egr1-activated transcription plays a key role in macrophage polarization: A multiomics sequencing study with partial validation. International Immunopharmacology, 99, 108072.
14. Song, S., Tang, Y., & Qin, R. (2025). Synthetic Data Matters: Re-training with Geo-typical Synthetic Labels for Building Detection. IEEE Transactions on Geoscience and Remote Sensing.
15. Guo, Y., Sekiguchi, Y., Zeng, W., Ebihara, S., Owaki, D., & Hayashibe, M. (2025). Physics-informed learning framework for lower limb kinematic prediction with sparse sensors and its application in chronic stroke. IEEE Transactions on Neural Systems and Rehabilitation Engineering.
16. Zhao, Y., Zhu, Z., Yang, S., & Li, W. (2026). YeastSAM: A Deep Learning Model for Accurate Segmentation of Yeast Cells. Molecular Biology of the Cell, mbc-E25.
17. De Fauw, J., Ledsam, J. R., Romera-Paredes, B., Nikolov, S., Tomasev, N., Blackwell, S., et al. (2018). Clinically applicable deep learning for diagnosis and referral in retinal disease. Nature Medicine, 24(9), 1342–1350.
18. Xu, H., Ge, J., & Ou, Y. (2026). DentaScope-AI: A Real-Time k-mer–Driven CNN-Attention Framework for Rapid Pathogen Identification and Virulence Prediction in Endodontic Infections Using Nanopore Sequencing. Frontiers in Healthcare Technology, 3(1), 112-122.
19. Cai, Lie, et al. "Deep learning model for breast shear wave elastography to improve breast cancer diagnosis (INSPiRED 006): An international, multicenter analysis." Journal of Clinical Oncology 43.32 (2025): 3482-3493.
20. Ren, Y., Wu, D., & Lopez–De Fede, A. (2022, June). Identification and Prediction of Low-Birthweight Baby Outcomes and Mom Risk Factors. In 2022 IEEE 10th International Conference on Healthcare Informatics (ICHI) (pp. 01-02). IEEE.
21. Yao, Y., Li, Y., Zheng, S., & Zhu, T. (2025). Deep Learning Models for Multi-class Pneumonia Detection in Chest X-rays: A Comparative Study of VGG16, MobileNet, and ResNet152. Computer Simulation in Application, 3(1), 66-73.
22. Yao, Y., Li, Y., Zheng, S., & Zhu, T. (2025). MSA-TransUNet: A Multi-Scale Attention Enhanced Transformer-UNet Architecture for Accurate Vessel Segmentation and Visualization in Medical Imaging. Journal of Medicine and Life Sciences, 1(4), 48-57.
23. Dong, J., Koniusz, P., Feng, L., Zhang, Y., Zhu, H., Liu, W., ... & Ong, Y. S. (2025). Robustifying zero-shot vision language models by subspaces alignment. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 21037-21047).
24. Bagheri, Mahdi, et al. "An MRI Atlas of the Human Fetal Brain: Reference and Segmentation Tools for Fetal Brain MRI Analysis." ArXiv (2025): arXiv-2508.
25. Dong, J., Chen, J., Xie, X., Lai, J., & Chen, H. (2024). Survey on adversarial attack and defense for medical image analysis: Methods and challenges. ACM Computing Surveys, 57(3), 1-38.
26. Qu, W., Wang, J., Gong, Y., Huang, X., & Xiao, L. (2025). An end-to-end robust point cloud semantic segmentation network with single-step conditional diffusion models. In Proceedings of the Computer Vision and Pattern Recognition Conference (pp. 27325-27335).
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Seojoon Lee, Adrian Benjamin, Yuji Saito (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
All articles published in the International Journal of Machine Learning and Engineering (IJMLE) are licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).
This license permits unrestricted use, distribution, reproduction, adaptation, and commercial use in any medium or format, provided the original author(s) and the source are properly credited.
Authors retain the copyright of their work while granting the journal the right of first publication.
To view a copy of this license, please visit:
https://creativecommons.org/licenses/by/4.0/
