OncoGPT: Enhancing oncological decision-making with ASCO guideline–trained AI.

e13607 Background: General-purpose language models like ChatGPT, while proficient in a wide array of topics, often lack the specialized knowledge and precision required for specific fields such as oncology. The American Society of Clinical Oncology (ASCO) guidelines offer a comprehensive resource for oncology professionals, yet sifting through extensive documents for specific information can be daunting. To bridge this gap, we introduce OncoGPT, a specialized GPT model, tailored to provide precise and guideline-adherent oncological advice. Methods: We leveraged GPT-4's advanced language understanding capabilities as the foundation for OncoGPT. This customization involved an extensive training regimen using the complete text of the ASCO clinical guidelines. The training process was meticulously designed to ensure that OncoGPT not only assimilates the detailed information within these guidelines but also understands the context and intricacies of oncological care. Additionally, we created disease-specific versions of OncoGPT, each fine-tuned with guideline subsets pertinent to particular oncological conditions. Further, the training was augmented with specific prompts to foster kind, sincere, and respectful interactions, especially for patient-oriented responses. The model was designed to discern whether the user is a healthcare professional or a patient and adjust the complexity of its responses accordingly. For patients, additional emphasis was placed on delivering information in a kind, responsible, and empathetic manner. Results: OncoGPT's performance was rigorously evaluated for accuracy and reasoning against real-world clinical scenarios. The responses generated by the model were meticulously reviewed by an experienced oncologist to confirm adherence to ASCO guidelines and clinical relevance. This assessment revealed a high level of precision and contextual understanding in OncoGPT's outputs, demonstrating its potential as a valuable tool for oncology practitioners. Importantly, it showed a remarkable ability to tailor its communication style based on the user's background, providing simplified explanations for patients and more detailed responses for professionals. An example of a specific OncoGPT: Breast Cancer OncoGPT: https://chat.openai.com/g/g-VcuECT3jg-oncogpt-breast-cancer Conclusions: OncoGPT represents a significant advancement in the application of AI for medical decision-making. By integrating the depth of ASCO guidelines into a responsive and accurate AI model, OncoGPT offers oncology professionals a powerful tool for accessible, guideline-compliant advice. This development marks a crucial step towards integrating AI more profoundly into specialized medical fields, enhancing both the efficiency and accuracy of clinical decision-making.