ICON: a computer-based approach to differential diagnosis in radiology

Area of Science:

• Radiology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Differential diagnosis of lung disease on chest radiographs is complex.
• Lymphoproliferative disorders present unique challenges in radiographic interpretation.
• Existing diagnostic tools may lack comprehensive evidence-based support.

Purpose of the Study:

• To develop and evaluate ICON, a computer-based expert system.
• To assist radiologists in the differential diagnosis of lung disease in patients with lymphoproliferative disorders.
• To provide an evidence-based critique of proposed diagnoses.

Main Methods:

• Development of an expert system (ICON) utilizing a knowledge base of lung diseases and lymphoproliferative disorders.
• System interaction involves radiologists proposing a diagnosis.
• ICON analyzes evidence supporting proposed and competing diagnoses.

Main Results:

• ICON provides diagnostic support by generating an English prose critique.
• The system integrates computer recall of information with human reasoning.
• Focuses on lung disease interpretation in the context of lymphoproliferative disorders.

Conclusions:

• ICON serves as a valuable tool for radiologists in differential diagnosis.
• The system enhances diagnostic reasoning by presenting evidence-based critiques.
• This approach combines computational power with clinical expertise for improved lung disease diagnosis.