Automated Pathologic TN Classification Prediction and Rationale Generation From Lung Cancer Surgical Pathology Reports Using a Large Language Model Fine-Tuned With Chain-of-Thought: Algorithm Development and Validation Study
2024

Automated Lung Cancer Staging Using Language Models

Sample size: 3216 publication Evidence: high

Author Information

Author(s): Castonguay Alexandre, Lampridis Savvas, Kim Sanghwan, Jang Sowon, Kim Borham, Sunwoo Leonard, Kim Seok, Chung Jin-Haeng, Nam Sejin, Cho Hyeongmin, Lee Donghyoung, Lee Keehyuck, Yoo Sooyoung

Primary Institution: Seoul National University Bundang Hospital

Hypothesis

Can fine-tuned large language models accurately classify pathologic TN stages and generate rationales from lung cancer surgical pathology reports?

Conclusion

The study demonstrates that fine-tuned generative language models can enhance and automate TN classification in lung cancer, improving accuracy and efficiency.

Supporting Evidence

  • The Orca2_13b model achieved the highest performance with an EMR of 0.934 and an SMR of 0.864.
  • The Orca2_7b model also demonstrated strong performance with an EMR of 0.914.
  • Performance was assessed using exact match ratio (EMR) and semantic match ratio (SMR) as evaluation metrics.

Takeaway

This study shows that computers can help doctors quickly and accurately figure out how advanced lung cancer is by reading pathology reports.

Methodology

The study evaluated 6 open-source large language models on 3216 deidentified lung cancer surgical pathology reports, assessing their performance using exact match ratio (EMR) and semantic match ratio (SMR).

Limitations

The study was limited to data from a single tertiary hospital, which may affect the generalizability of the results.

Digital Object Identifier (DOI)

10.2196/67056

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication