Deep learning using contrast-enhanced ultrasound images to predict the nuclear grade of clear cell renal cell carcinoma.

PURPOSE: To assess the effectiveness of a deep learning model using contrastenhanced ultrasound (CEUS) images in distinguishing between low-grade (grade I and II) and high-grade (grade III and IV) clear cell renal cell carcinoma (ccRCC). METHODS: A retrospective study was conducted using CEUS images of 177 Fuhrmangraded ccRCCs (93 low-grade and 84 high-grade) from May 2017 to December 2020. A total of 6412 CEUS images were captured from the videos and normalized for subsequent analysis. A deep learning model using the RepVGG architecture was proposed to differentiate between low-grade and high-grade ccRCC. The model's performance was evaluated based on sensitivity, specificity, positive predictive value, negative predictive value and area under the receiver operating characteristic curve (AUC). Class activation mapping (CAM) was used to visualize the specific areas that contribute to the model's predictions. RESULTS: For discriminating high-grade ccRCC from low-grade, the deep learning model achieved a sensitivity of 74.8%, specificity of 79.1%, accuracy of 77.0%, and an AUC of 0.852 in the test set. CONCLUSION: The deep learning model based on CEUS images can accurately differentiate between low-grade and high-grade ccRCC in a non-invasive manner.

The kinetic quantitative characteristics of non-mass breast lesions with contrast-enhanced ultrasound: a prospective study.

OBJECTIVE: To characterize non-mass breast lesions (NML) quantitatively by contrast-enhanced ultrasound (CEUS) and to evaluate its additional diagnostic value based on the Breast Imaging Reporting and Data System (BI-RADS) categories. METHODS: A prospective study was performed among consecutive patients with NMLs. All lesions were examined by grayscale ultrasound and CEUS and diagnosed on pathology. Standard mammograms were obtained in the patients over 30 years old. Three independent radiologists assessed the features on grayscale ultrasound and mammograms and classified NMLs according to BI-RADS categories. Combined with the quantitative analysis in CEUS, the BI-RADS categories were reassessed, and the sensitivity, specificity, positive-predictive value, negative-predictive value and area under the receiver operating characteristic curve (AUC) were calculated for the evaluation of the diagnostic performance. RESULTS: 30 benign and 24 malignant NMLs were finally enrolled in this study, with ductal carcinoma in situ being the majority of malignant (15/24). Average contrast signal intensity (AI), wash-in rate (WiR) and enhancement intensity at 40 s (I40) were found to be the most efficient kinetic parameters to diagnose malignant NMLs. Combined with the cut-off values of 205.2 for AI, 127.8 for WiR and 136.4 for I40, the diagnostic accuracy was improved (AUC = 0.904), with the sensitivity of 95.8% and the specificity of 70.0%. CONCLUSION: The results suggested that hyperenhancement and rapid wash-in and wash-out are the characteristics of malignant NMLs. The kinetic analysis using CEUS can reflect hypervascular nature of malignant NMLs, thus improving the diagnostic performance combined with grayscale ultrasound. ADVANCES IN KNOWLEDGE: In this study, we quantified the enhancement characteristics of non-mass breast lesions with CEUS. We revealed that the combination of CEUS and conventional ultrasound provided higher sensitivity for diagnosing malignant NMLs.