Clinical Application of Computer-Aided Diagnosis System in Breast Ultrasound: A Prospective Multicenter Study

Area of Science:

• Medical Imaging
• Radiology
• Oncology

Background:

• Ultrasound exhibits high sensitivity but low specificity for breast lesions, leading to unnecessary biopsies.
• Computer-aided diagnosis (CAD) aims to improve diagnostic accuracy and reduce invasive procedures.

Purpose of the Study:

• To evaluate the diagnostic performance of the S-Detect CAD system in differentiating breast lesions.
• To assess the system's ability to reduce unnecessary breast biopsies in non-university hospitals in less-developed regions of China.

Main Methods:

• A prospective multicenter study involving 8 hospitals.
• Analysis of ultrasound images, CAD data, and BI-RADS classifications.
• Comparison of diagnostic accuracy, sensitivity, specificity, PPV, NPV, and AUC between CAD-assisted and unassisted radiologists.

Main Results:

• CAD combined with less-experienced radiologists showed superior AUC (0.878 vs 0.712) and specificity (81.3% vs 44.6%) compared to unassisted radiologists.
• No significant difference in AUC or specificity was observed between experienced radiologists and less-experienced radiologists using CAD.
• CAD significantly decreased the biopsy rate (25.6% vs 100.0%) and increased the malignancy rate of biopsies (43.9% vs 15.0%) for less-experienced radiologists.

Conclusions:

• The CAD system serves as an effective auxiliary tool for differentiating breast lesions.
• It helps reduce unnecessary biopsies for radiologists in less-developed regions of China.
• CAD enhances diagnostic performance, particularly for less-experienced practitioners.