Ha, Eun Ju1
1 Department of Radiology, Ajou University School of Medicine, Wonchon-Dong, Yeongtong-Gu, Korea
Purpose: To prospectively evaluate the diagnostic performance of a computer-aided diagnosis (CAD) system (S-Detect for ThyroidTM) for thyroid US in the differential diagnosis of thyroid nodules and to determine interobserver agreement between the CAD and experienced radiologist.
Materials & Methods: We consecutively enrolled patients with thyroid nodules with decisive diagnosis whether benign or malignant, from June 2016 and July 2016. An experienced radiologist reviewed US images characteristics of thyroid nodules in a prospective design, and a CAD system additionally provided nodule diagnosis whether benign or malignant. We compared the diagnostic performance of experienced radiologist, CAD system, and a combined assessment.
Results: A total of 117 thyroid nodules from 50 consecutive patients were included. The mean size of nodules was 1.5 ± 1.1 cm and final diagnoses were 67 (57.3%) benign nodules and 50 (42.7%) malignant nodules. The CAD system showed similar sensitivity and specificity compared with the experienced radiologist (sensitivity: 80.0% versus 87.0%, P= 0.754; specificity: 88.1% versus 95.5%, P= 0.180). The diagnostic accuracy was not different between the CAD and experienced radiologist (84.6% versus 90.6%, P= 0.646). The CAD system assisted radiologist reached the higher diagnostic sensitivity when compared to the radiologist only (91.8% versus 87.0%, P= 0.031). The interobserver agreement between the experienced radiologist and CAD system were substantial agreement for a final diagnosis (kappa=0.661).
Conclusion: The diagnostic performance of CAD system for differentiating thyroid nodule was as good as that of the experienced radiologist with a substantial agreement. The CAD system assisted radiologist could achieve the highest diagnostic sensitivity for thyroid cancer.
- Chang Y, Paul AK, Kim N, et al. Computer-aided diagnosis for classifying benign versus malignant thyroid nodules based on ultrasound images: A comparison with radiologist-based assessments. Medical physics 2016;43:554-567.
- Choi YJ, Baek JH, Park HS, et al. A computer-aided diagnosis system using artifical intelligence for the diagnosis and characterization of thyroid nodules on ultrasound: initial clinical assessment. Thyroid. 2017 Jan 10. doi: 10.1089/thy.2016.0372. [Epub ahead of print]