Clinical Report: Utilizing Large Language Models and Machine Learning to Enhance Chemotherapy Choices in Breast Cancer
Overview
This study presents a hybrid framework that integrates machine learning and large language models to improve chemotherapy decision-making in breast cancer. The Random Forest classifier demonstrated high predictive accuracy, while GPT-4 provided clinically relevant rationales for treatment recommendations.
Background
Breast cancer remains a leading cause of cancer-related mortality among women, with chemotherapy decisions often complicated by biological variability and inconsistent clinical practices. Current methods for determining chemotherapy eligibility can lead to both overtreatment and undertreatment.
Data Highlights
Model
AUC
Cohen's κ
Random Forest
0.91
-
GPT-4
-
0.13
Key Findings
The Random Forest classifier achieved an AUC of 0.91, outperforming baseline models.
Causal analysis revealed heterogeneous treatment benefits, indicating potential patient groups for chemotherapy deprioritization.
GPT-4 demonstrated moderate agreement with the Random Forest model in generating treatment rationales.
Uplift-based machine learning policies outperformed traditional treat-all and treat-none strategies.
GPT-4 enhanced interpretability through rationale-driven explanations.
Clinical Implications
The proposed framework leverages machine learning and causal analysis to support chemotherapy recommendations.
Conclusion
This study integrates advanced computational methods to enhance chemotherapy decision-making in breast cancer.
by Md Serajun Nabi, Dema Yuden, Thinley Yeshey Choden, S. M. Asiful Islam Saky, Hasanul Bannah, Md Sabbir Hossen, Mohammad Faizal Ahmad Fauzi, Hezerul Bin Abdul Karim
