Applying Artificial Intelligence to Address the Knowledge Gaps in Cancer Care

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Rapid scientific advancements necessitate improved methods for integrating evidence-based care into community settings.
• A gap exists between current medical practices and the latest scientific possibilities, particularly in oncology.

Purpose of the Study:

• To research and develop an artificial intelligence (AI) application for real-time, patient-specific decision support in oncology.
• To create a tool that bridges the gap between scientific discovery and clinical practice.

Main Methods:

• Developed the Oncology Expert Advisor (OEA) application with three core functions: patient history summarization, treatment recommendation, and management advisory.
• Utilized machine-learning algorithms trained on retrospective patient data (approx. 1,000 patients) and the Medline database (over 23 million abstracts).

Main Results:

• OEA demonstrated high accuracy in extracting non-time-dependent clinical concepts (90%-96% F1 score) and moderate accuracy for time-dependent concepts (63%-65% F1 score).
• The system achieved high recall (99.9%) and 88% precision in suggesting approved therapy options and high recall (97.9%) and precision (96.9%) in screening for clinical trials.

Conclusions:

• The AI application is technically feasible for constructing patient profiles and suggesting evidence-based treatments and clinical trials.
• Collaboration between clinical and AI domains, with clinical expertise integrated throughout the development process, is crucial for successful implementation.