ReportTutor - an intelligent tutoring system that uses a natural language interface.

Area of Science:

• Medical Education
• Digital Pathology
• Artificial Intelligence in Medicine

Background:

• Intelligent Tutoring Systems (ITS) are valuable tools for medical education.
• Visual diagnosis in pathology requires accurate and complete reporting.
• Existing ITS often lack integration with visual diagnostic tools.

Purpose of the Study:

• To introduce ReportTutor, an ITS designed for visual diagnosis in pathology.
• To evaluate the system's ability to provide feedback on student diagnostic reports.
• To assess the system's performance in identifying errors and distinguishing correct from incorrect report elements.

Main Methods:

• ReportTutor integrates a virtual microscope with a natural language interface.
• The system monitors user actions and text input for real-time feedback.
• Utilizes MMTx, NCI Metathesaurus, and a cancer-specific ontology for domain knowledge and evaluation.
• Data collected from 4 pathology residents compared to expert dermatopathologist reports.

Main Results:

• Analysis of student reports identified error rates and distributions.
• The system demonstrated performance in identifying features within student reports.
• The system accurately distinguished between correct and incorrect report elements.

Conclusions:

• ReportTutor shows promise as an ITS for pathology training.
• The system's feedback mechanism can guide students towards more accurate and complete diagnostic reports.
• Further evaluation is needed to fully establish its efficacy in improving diagnostic skills.