Gemini Demonstrates Just-in-Time Oncology Analysis From Text
This study (May et al., 2025) evaluated Gemini 2.5 Pro on a synthetic dataset of 240 unstructured stage IV NSCLC clinical letters to test a just-in-time (JIT) oncology analysis workflow. The LLM achieved >98% extraction accuracy, cut processing time dramatically (3.7 versus 133.8 minutes), and generated executable R code producing Kaplan-Meier results indistinguishable from ground truth; results remain limited to synthetic data.
Key Points
- 1Achieved >98% multiparameter extraction accuracy across 240 synthetic NSCLC letters.
- 2Reduced processing time (3.7 min vs 133.8 min), enabling faster cohort-level analyses.
- 3Automated end-to-end survival analysis generating executable R code with matching statistical results.
Scoring Rationale
Demonstrates practical, high-fidelity LLM analytics with executable outputs; limited by synthetic datasets and need for real-world validation.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
