JOURNAL CLUB: Computer-Aided Detection of Lung Nodules on CT With a Computerized Pulmonary Vessel Suppressed Function

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Early and accurate detection of lung nodules on chest CT is crucial for patient outcomes.
• Computer-aided detection (CADe) systems aim to improve diagnostic performance.
• Pulmonary vessel suppression can enhance nodule visualization.

Purpose of the Study:

• To evaluate the effectiveness of a novel pulmonary vessel image-suppressed function with an integrated computer-aided detection (VIS/CADe) system.
• To assess the impact of the VIS/CADe system on radiologists' performance in detecting actionable nodules on chest CT.
• To measure changes in interpretation time with the use of the VIS/CADe system.

Main Methods:

• Development of a novel VIS/CADe system for thoracic CT interpretation.
• A comparative reader study involving 12 radiologists interpreting 324 cases (33% with nodules).
• Localization ROC (LROC) analysis to evaluate detection performance with and without the VIS/CADe system.

Main Results:

• The VIS/CADe system demonstrated high standalone detection rates for malignant (89.5%) and small (≥5 mm, 82.0%) nodules.
• Radiologists' lung cancer detection significantly improved with VIS/CADe (LROC-AUC 0.773 vs. 0.633, p < 0.01).
• Actionable nodule detection improved (LROC-AUC 0.692 vs. 0.584, p < 0.01), with a 26% reduction in interpretation time.

Conclusions:

• The VIS/CADe system significantly enhances radiologists' detection of lung cancers and actionable nodules.
• While specificity slightly decreased, the overall diagnostic improvement and efficiency gains are substantial.
• The VIS/CADe system shows strong potential to aid radiologists in identifying critical findings on thoracic CT scans.