Convolutional Neural Network to Stratify the Malignancy Risk of Thyroid Nodules: Diagnostic Performance Compared with the American College of Radiology Thyroid Imaging Reporting and Data System Implemented by Experienced Radiologists.

BACKGROUND AND PURPOSE: Comparison of the diagnostic performance for thyroid cancer on ultrasound between a convolutional neural network and visual assessment by radiologists has been inconsistent. Thus, we aimed to evaluate the diagnostic performance of the convolutional neural network compared with the American College of Radiology Thyroid Imaging Reporting and Data System (TI-RADS) for the diagnosis of thyroid cancer using ultrasound images. MATERIALS AND METHODS: From March 2019 to September 2019, seven hundred sixty thyroid nodules (≥10 mm) in 757 patients were diagnosed as benign or malignant through fine-needle aspiration, core needle biopsy, or an operation. Experienced radiologists assessed the sonographic descriptors of the nodules, and 1 of 5 American College of Radiology TI-RADS categories was assigned. The convolutional neural network provided malignancy risk percentages for nodules based on sonographic images. Sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were calculated with cutoff values using the Youden index and compared between the convolutional neural network and the American College of Radiology TI-RADS. Areas under the receiver operating characteristic curve were also compared. RESULTS: Of 760 nodules, 176 (23.2%) were malignant. At an optimal threshold derived from the Youden index, sensitivity and negative predictive values were higher with the convolutional neural network than with the American College of Radiology TI-RADS (81.8% versus 73.9%, P = .009; 94.0% versus 92.2%, P = .046). Specificity, accuracy, and positive predictive values were lower with the convolutional neural network than with the American College of Radiology TI-RADS (86.1% versus 93.7%, P < .001; 85.1% versus 89.1%, P = .003; and 64.0% versus 77.8%, P < .001). The area under the curve of the convolutional neural network was higher than that of the American College of Radiology TI-RADS (0.917 versus 0.891, P = .017). CONCLUSIONS: The convolutional neural network provided diagnostic performance comparable with that of the American College of Radiology TI-RADS categories assigned by experienced radiologists.

Additional malignant breast lesions detected on second-look US after breast MRI vs. additional malignant lesions detected on initial US in breast cancer patients: comparison of US characteristics.

PURPOSE: The purpose of our study was to review and compare the US findings of synchronous malignant breast lesions other than the index cancer additionally detected on second-look US with those detected on initial US, and therefore to determine differing characteristics that may aid in diagnosis and essentially improve the performance of the initial US examination. MATERIALS AND METHODS: A retrospective review of 39 mammographically occult synchronous malignant lesions other than the index cancer from 38 patients was performed (21 lesions: detected on second-look US, 18 lesions: detected on initial US). All patients underwent initial mammography, bilateral whole breast US (BWBU) and breast MRI, and all lesions were confirmed pathologically by biopsy or preoperative localization. RESULTS: Additional malignant breast lesions detected on both initial US and second-look US tended to be subtle and often did not show classic malignant findings. Second-look US lesions (median, 7.0 mm; range, 3 - 22 mm) tended to be smaller than initial US lesions (median, 9.0 mm; range 3 - 45 mm), although the difference was not statistically significant (p = 0.134). Second-look US lesions also showed no posterior acoustic features (p = 0.037) and a significantly higher proportion of lesions with circumscribed or indistinct margins compared to initial US lesions (p = 0.042). Second-look US lesions were significantly subareolar or relatively far (> 5 cm) from the nipple than initial US lesions (p = 0.048). CONCLUSION: Second-look US lesions showed more subtle findings of posterior acoustic features and margins, and tended to be subareolar or relatively far from the nipple compared to initial US lesions. However, both groups showed subtle US findings and there was no significant difference in other features.