To assess the value of a computer aided diagnosis (CAD) system based on the classification of thyroid imaging reporting and data system (TI-RADS) in the ultrasound diagnosis of thyroid cancer.

From October 2018 to March 2019, 400 cases of thyroid nodules were collected at five domestic hospitals for a multi-center retrospective study. A CAD system based on the TI-RADS classification developed by Peking University was used. With and without the aid of the CAD system, 11 radiologists with different experiences (4 junior radiologists, 4 intermediate-level radiologists, and 3 senior radiologists) diagnosed thyroid cancer, respectively. The area under the receiver operating characteristic (ROC) curve (AUC) was compared using the DeLong method, the sensitivity and specificity were compared using the paired chi-square test, and the reading time was compared using the paired t test.

With the aid of the CAD system, the average AUC of all radiologists for diagnosing thyroid cancer was significantly higher than that without the aid of the CAD system [0.800 (0.788-0.812) vs 0.848 (0.837-0.858), P＜0.001]; the sensitivity significantly increased from 73.8% (71.9%-75.6%) to 82.7% (81.0%-84.3%, χ²=9.870, P＜0.001), and the specificity increased from 86.2% (84.7%-87.6%) to 86.9% (85.4%-88.3%), but there was no statistical difference (χ²=0.021, P=0.379). When performing subgroup analysis, with the aid of the CAD system, the AUC values of junior and intermediate-level radiologists for diagnosing thyroid cancer were significantly higher than those without the aid of the CAD system (junior radiologists: 0.740 vs 0.840, P＜0.001; intermediate-level radiologists: 0.814 vs 0.848, P=0.001), but the AUC value of senior radiologists was not significantly different from that without the aid of the CAD system (0.861 vs 0.859, P=0.861). With the aid of the CAD system, the reading time of all radiologists was significantly shortened compared with that without the aid of the CAD system (22.7±9.6 vs 20.2±8.2; P＜0.001).

The CAD system is helpful in the diagnosis of thyroid cancer by both junior radiologists and intermediate-level radiologists, and can reduce the time to achieve diagnosis.