Development and evaluation of a computer-based decision support system for diffuse lung diseases at high-resolution...

Area of Science:

• Radiology
• Medical Informatics
• Pulmonology

Background:

• High-resolution computed tomography (HRCT) is crucial for diagnosing diffuse and interstitial lung diseases.
• Accurate interpretation of HRCT patterns is essential for effective patient management.
• Challenges exist in consistently identifying and differentiating various lung pathologies on HRCT.

Purpose of the Study:

• To evaluate a novel computer-based decision support system (CDSS) for aiding HRCT diagnosis of diffuse lung diseases.
• To assess the impact of a CDSS on the diagnostic performance of radiology trainees.
• To determine if a structured, illustrative approach enhances the interpretation of HRCT findings.

Main Methods:

• A CDSS was developed featuring approximately 100 common HRCT signs, patterns, and associated pathologies.
• The system facilitates structured evaluation by allowing selection of CT patterns to generate differential diagnoses.
• Three blinded radiology residents evaluated 40 lung disease cases twice: once unaided and once with the CDSS.

Main Results:

• The CDSS improved correct pattern classification from 90.3% to 96.8% (P < 0.01).
• Correct diagnosis rates increased significantly with CDSS use (81.7% vs. 64.2%, P < 0.01).
• Differential diagnosis accuracy substantially improved from 38.3% to 89.2% (P < 0.01) with the CDSS.

Conclusions:

• A CDSS significantly enhances the ability of trainees to characterize and diagnose diffuse lung diseases on HRCT.
• Structured evaluation supported by graphical illustrations and pattern explanations is key to improved diagnostic performance.
• The developed CDSS shows promise as a valuable tool for medical education and clinical practice in thoracic radiology.