Ultrasound of Thyroid Nodules and the Thyroid Imaging Reporting and Data System.

Benign or malignant thyroid nodules are common in adults. Fine needle aspiration biopsy is the gold standard for diagnosis. Most thyroid nodules are benign. Ultrasound imaging is the optimal noninvasive imaging modality to determine which nodules demonstrate malignant features. The American College of Radiology Thyroid Imaging Reporting and Data System committee published a standardized approach to classifying nodules on ultrasound. The ultrasound features in this system are categorized as benign, minimally suspicious, moderately suspicious, or highly suspicious for malignancy. Applying the Thyroid Imaging Reporting and Data System results in a meaningful decrease in the number of thyroid nodules biopsied.

Peripheral Thyroid Nodule Calcifications on Sonography: Evaluation of Malignant Potential.

OBJECTIVE. The purpose of this study is to assess the association of thyroid cancer with sonographic features of peripheral calcifications. MATERIALS AND METHODS. We retrospectively reviewed patients who had a total of 97 thyroid nodules with peripheral calcifications who underwent ultrasound-guided fine-needle aspiration from 2008 to 2018. Three board-certified radiologists evaluated the nodules for features of peripheral calcifications: the percentage of the nodule involved by peripheral calcifications, whether the calcifications were continuous or discontinuous, the visibility of internal components of the nodule, and the presence of extrusion of soft tissue beyond the calcifications. The correlation of peripheral calcification parameters with the rate of thyroid nodule malignancy was evaluated. In addition, the interobserver agreement between readers was assessed with Cohen kappa coefficient. RESULTS. Of the 97 nodules with peripheral calcifications, 27% (n = 26) were found to be malignant on biopsy. The continuity of peripheral calcifications, visibility of internal components, and extrusion of soft tissue beyond the calcification rim showed no significant association with benign or malignant nodules. Readers had good agreement on peripheral calcification continuity (κ = 0.63; 95% CI, 0.53-0.73) and moderate agreement on internal component visibility (κ = 0.43; 95% CI, 0.35-0.51) and percentage of the nodule involved by rim calcifications (κ = 0.52; 95% CI, 0.44-0.59). There was fair agreement for extranodular soft-tissue extrusion (κ = 0.32, 95% CI, 0.24-0.39). CONCLUSION. Peripheral rim calcifications are highly associated with malignancy. However, specific peripheral rim calcification features do not aid in distinguishing benign from malignant nodules, which may in part be caused by high interobserver variability.

Transfer Learning with Convolutional Neural Networks for Classification of Abdominal Ultrasound Images.

The purpose of this study is to evaluate transfer learning with deep convolutional neural networks for the classification of abdominal ultrasound images. Grayscale images from 185 consecutive clinical abdominal ultrasound studies were categorized into 11 categories based on the text annotation specified by the technologist for the image. Cropped images were rescaled to 256 × 256 resolution and randomized, with 4094 images from 136 studies constituting the training set, and 1423 images from 49 studies constituting the test set. The fully connected layers of two convolutional neural networks based on CaffeNet and VGGNet, previously trained on the 2012 Large Scale Visual Recognition Challenge data set, were retrained on the training set. Weights in the convolutional layers of each network were frozen to serve as fixed feature extractors. Accuracy on the test set was evaluated for each network. A radiologist experienced in abdominal ultrasound also independently classified the images in the test set into the same 11 categories. The CaffeNet network classified 77.3% of the test set images accurately (1100/1423 images), with a top-2 accuracy of 90.4% (1287/1423 images). The larger VGGNet network classified 77.9% of the test set accurately (1109/1423 images), with a top-2 accuracy of VGGNet was 89.7% (1276/1423 images). The radiologist classified 71.7% of the test set images correctly (1020/1423 images). The differences in classification accuracies between both neural networks and the radiologist were statistically significant (p < 0.001). The results demonstrate that transfer learning with convolutional neural networks may be used to construct effective classifiers for abdominal ultrasound images.