Endocrine Abstracts

Aim: To evaluate the diagnostic effectiveness of the first commercially available computer-aided diagnostic (CAD) imaging analysis system.

Methods: Ultrasound images of 300 thyroid nodules (135 malignant) acquired prior to surgical treatment were retrospectively reviewed by a thyroid expert, and then the classification of each image was compared to the classification rendered by an image analysis program (AmCAD-UT^®). The American Thyroid Association (ATA), the European Thyroid Imaging Reporting and Data System (EU-TIRADS), and the American Association of Clinical Endocrinologists (AACE/ACE/AME) classification systems were used for risk stratification.

Results: Diagnostic performance of the thyroid expert using the ATA system was: 87.0% sensitivity, 91.2% specificity, 90.5% positive predictive value (PPV) and 90.9% negative predictive value (NPV). Compared to the clinical expert, sensitivity of the CAD program based on the three reporting systems was similar, whereas specificity and PPV were lower. Regarding NPV, results for the clinician did not differ from the CAD program when it applied the ATA classification system (90.9% vs. 86.3%, P=0.07), but when it applied the EU-TIRADS and AACE/ACE/AME classifications it performed less well (82.6% and 80.8%, respectively; P=0.01 in both cases). The area under the receiver operating characteristic curve (AUC) was 0.88 for the expert clinician and 0.72 for the CAD based on ATA reporting system.

Conclusions: The CAD ultrasound image analysis program evaluated in this study is a useful tool for risk stratification of thyroid nodules, but does not perform better than a clinical expert.