Utility of Artificial Intelligence for Decision Making in Thoracic Multidisciplinary Tumor Boards

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Multidisciplinary tumor boards (MTBs) are crucial for complex cancer treatment decisions.
• Integrating advanced AI tools like Natural Language Processing (NLP) could potentially enhance MTB decision-making processes.

Purpose of the Study:

• To evaluate the efficacy of OpenAI's NLP (GPT 3.5 turbo chat) in supporting thoracic MTBs for non-small-cell lung cancer (NSCLC) treatment recommendations.
• To compare AI-generated recommendations with actual MTB decisions based on clinical guidelines.

Main Methods:

• Retrospective comparative study of 52 NSCLC patients presented to a thoracic MTB.
• AI (GPT 3.5 turbo chat) was provided with patient data and clinical guidelines (SEPAR) to generate treatment recommendations.
• Concordance analysis using the Kappa coefficient to assess agreement between AI and MTB decisions.

Main Results:

• The AI achieved an overall concordance rate of 76% with the MTB decisions.
• A Kappa index of 0.59 indicated moderate agreement.
• High consistency (92.3%) and replicability were observed for AI's surgical recommendations.

Conclusions:

• AI, specifically NLP, shows promise as a supportive tool for MTB decision-making in thoracic oncology.
• Further research is warranted to explore the full potential and integration of AI in clinical practice.
• AI can aid in standardizing and potentially improving the consistency of cancer treatment recommendations.