To assess the clinical value of computer-aided detection and diagnosis (CAD) system in evaluating the necessity of fine needle aspiration cytology (FNAC) in thyroid nodules.

A total of 106 thyroid nodules were prospectively included in this study. The CAD system, a senior radiologist, and a junior radiologist assessed and determined the necessity of FNAC in thyroid nodules independently. Every nodule was stratified according to the American Thyroid Association guidelines. If the FNAC pathological result was malignancy, the FNAC recommendation was considered meaningful, otherwise the FNAC recommendation was considered meaningless. The diagnostic sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the senior radiologist, the junior radiologist, and the CAD system for evaluating thyroid nodules were compared. Thyroid nodules were divided into three groups according to nodule size: 1.05 cm ~ ＜1.5 cm group, 1.5 cm ~ ＜2.0 cm group, and ≥ 2.0 cm group. The effect of nodule size on the efficacy of CAD software in evaluating thyroid nodule puncture was studied. The differences in sensitivity, specificity, and accuracy among different groups were compared. The McNemar test was used for comparing differences in this study.

The sensitivity of the CAD system was lower than that of the senior radiologist (71.2% vs 84.6%, P=0.039), but the specificity was higher than that of the senior radiologist (77.8% vs 61.1%, P=0.049). Both the sensitivity and specificity of the CAD were higher than those of the junior radiologist (71.2% vs 55.8%, P=0.021; 77.8% vs 64.8%, P=0.039). The accuracy of the CAD software was similar to that of the senior radiologist, but higher than that of the junior radiologist (74.5% vs 60.4%, P=0.001). The evaluation efficiency of the CAD system was related to the size of nodules. The sensitivity of the CAD in the 1.0 cm ~ ＜1.5 cm group was the highest. When the maximum diameter of nodules was ≥2.0 cm, the specificity and accuracy of the CAD were the highest (91.7% vs 80.0% vs 53.3%; 84.4% vs 81.6% vs 48.0%); the differences were statistically significant (P=0.023 and 0.002, respectively).

The CAD system can be used to determine the detect the necessity of FNAC for suspicious lesions. There is no significant difference in accuracy of assessme nt between the CAD system and the senior radiologist, while the accuracy of the CAD system is higher than that of the junior radiologist. Nodule size is a potential factor affecting evaluation performance of the CAD system.