Machine Learning Assisted Doppler Features for Enhancing Thyroid Cancer Diagnosis: A Multi-Cohort Study.

BACKGROUND: This pilot study aims at exploiting machine learning techniques to extract color Doppler ultrasound (CDUS) features and to build an artificial neural network (ANN) model based on these CDUS features for improving the diagnostic performance of thyroid cancer classification. METHODS: A total of 674 patients with 712 thyroid nodules (TNs) (512 from internal dataset and 200 from external dataset) were randomly selected in this retrospective study. We used ANN to build a model (TDUS-Net) for classifying malignant and benign TNs using both the automatically extracted quantitative CDUS features (whole ratio, intranodular ratio, peripheral ratio, and number of vessels) and gray-scale ultrasound (US) features defined by the American College of Radiology (ACR) Thyroid Imaging Reporting and Data System (TI-RADS). Then, we compared the diagnostic performance of the model, the performance of another ANN model based on the gray-scale US features alone (TUS-Net), and that of radiologists. RESULTS: The TDUS-Net (0.898, 95% CI: 0.868-0.922) achieved a higher area under the curve (AUC) than that of TUS-Net (0.881, 95% CI: 0.850-0.908) in the internal tests. Compared with radiologists, TDUS-Net (AUC: 0.925, 95% CI: 0.880-0.958) performed better than radiologists (AUC: 0.810, 95% CI: 0.749-0.862) in the external tests. CONCLUSIONS: Applying a machine learning model by combining both gray-scale US features and CDUS features can achieve comparable or even higher performance than radiologists in classifying TNs.

Added Value of Superb Microvascular Imaging and Virtual Touch Imaging Quantification in Assisting Thyroid Cancer Classification.

This prospective study determined the value added by superb microvascular imaging (SMI) and Virtual Touch imaging quantification (VTIQ) to conventional ultrasonography in classifying malignant and benign thyroid nodules. One hundred eighty-three thyroid nodules (TNs) in 120 patients (112 benign and 71 malignant TNs) were evaluated. SMI revealed noticeable variance between benign and malignant TNs (p < 0.001). Malignant nodules tended to have rich vascularity (grade 3: 38/71, 53.5%) compared with benign nodules (grade 3: 33/112, 29.5%). There is a statistically significant difference between malignant and benign TNs with respect to shear-wave speed (SWS) values (all p values <0.001). The SWS mean, maximum and ratio of malignant nodules were 3.97 ± 1.34, 4.79 ± 1.70 and 1.25 ± 0.39, respectively. The SWS mean, maximum and ratio of benign nodules were 2.65 ± 0.42, 2.97 ± 0.46 and 1.15 ± 0.35, respectively. With respect to area under the curve values, the combined use of SMI or VTIQ improved the diagnostic performance of classifying malignant and benign TNs compared with that of ultrasonography alone. The combination of three modalities achieved the greatest area under the curve values (0.9811, 95% confidence interval: 0.95529-1.000), followed by US + VTIQ (0.9747, 0.94543-1.000), US + SMI (0.9032, 0.85345-0.95391) and ultrasonography (0.8291, 0.76417-0.89403).

Respiratory Variations in Peak Peripheral Artery Velocities and Waveforms for Rapid Assessment of Fluid Responsiveness in Traumatic Shock Patients.

BACKGROUND This study aimed to assess the correlation between the variability of the end-inspiratory and end-expiratory blood flow waveform and fluid responsiveness (FR) in traumatic shock patients who underwent mechanical ventilation by evaluating peripheral arterial blood flow parameters. MATERIAL AND METHODS A cohort of 60 patients with traumatic shock requiring mechanical ventilation-controlled breathing received ultrasound examinations to assess the velocity of carotid artery (CA), femoral artery (FA) and brachial artery (BA). A rehydration test was performed in which of 250 mL of 0.9% saline was administered within 30 min between the first and second measurement of cardiac output by echocardiography. Then, all patients were divided into 2 groups, a responsive group (FR+) and a non-responsive group (FR-). The velocity of end-inspiratory and end-expiratory peripheral arterial blood flow of all patients was ultrasonically measured, and the variability were measured between end-inspiratory and end-expiratory. RESULTS The changes in the end-inspiratory and end-expiratory carotid artery blood flow velocity waveforms of the FR+ groups were significantly different from those of the FR- group (P<0.001). A statistically significant difference in ΔVmax (CA), ΔVmax (BA), and ΔVmax (FA) between these 2 groups was found (all P<0.001). The ROC curve showed that DVmax (CA) and ΔVmax (BA) were more sensitive values to predict FR compared to ΔVmax (FA). The sensitivity of ΔVmax (CA), ΔVmax (FA), and ΔVmax (BA) was 70.0%, 86.7%, and 93.3%, respectively. CONCLUSIONS The study showed that periodic velocity waveform changes in the end-inspiratory and end-expiratory peripheral arterial blood flow can be used for quick assessment of fluid responsiveness.

A comparative study on superb microvascular imaging and conventional ultrasonography in differentiating BI-RADS 4 breast lesions.

This prospective study aimed to explore the diagnostic value of superb microvascular imaging (SMI) in differentiating Breast Imaging Reporting and Data System (BI-RADS) 4 breast lesions compared with conventional ultrasonography (US). A total of 111 patients with 116 breast lesions underwent grayscale ultrasound (US), colour Doppler flow imaging (CDFI) and SMI breast imaging between February 2016 and May 2018. CDFI and SMI were performed to evaluate vascular quantity, morphology, and distribution characteristics. The detection of malignancy was compared between grayscale US alone, US + CDFI and US + SMI in terms of the BI-RADS stratification system. SMI was observed to be significantly more accurate in distinguishing malignant breast lesions (86.67%) compared with CDFI (80.00%) (P<0.001). Among malignant lesions, SMI detected 80.00% of those that contained ≥4 vessels, while CDFI only detected 56.67%. Penetrating and branching vessels were identified by SMI in 53.33% of malignant breast lesions and by CDFI in 10.00%. There was no significant difference in vascular distribution by SMI (P=0.094) and by CDFI (P=0.087). US + SMI was associated with higher sensitivity, specificity, and accuracy rates (86.67, 83.72 and 84.48%, respectively) compared with US + CDFI (80.00, 72.09 and 74.14%, respectively). The area under the curve values from receiver operating characteristic analysis of US + SMI, US + CDFI and US alone were 0.852 [95% confidence interval (CI): 0.768-0.936)] 0.760 (95% CI: 0.660-0.860), 0.698 (95% CI: 0.589-0.807), respectively (P<0.001). SMI yielded more detailed vascular information associated with malignant breast masses when compared with conventional US. Therefore, as an adjunct to grayscale, SMI exhibited a markedly improved diagnostic capability in distinguishing malignant and benign breast lesions, particularly those of BI-RADS category 4.

Evaluation of plasma cell mastitis with superb microvascular imaging.

BACKGROUND: Plasma cell mastitis (PCM), a common type of mastitis often mimics malignancy clinically and radiologically. OBJECTIVE: The study was designed to explore the diagnostic value of superb microvascular imaging (SMI) in differentiating PCM from malignant breast lesions. METHODS: A total of 95 breast lesions underwent conventional ultrasound (US) and SMI examination between May 2016 and April 2018. Vessels were detected in SMI in a quantitative manner. Blood flow parameters including systolic peak velocity (SPV), resistance index (RI), and pulsatility index (PI) were evaluated. We further assessed the diagnostic performances of US and US+SMI. RESULTS: The majority of PCM were in regular shape and displayed no calcification compared with malignant breast lesions. Regarding blood flow parameters, PCM obtained significantly lower mean value of RI and PI compared with malignant lesions (P < 0.05). The sensitivity, specificity and accuracy rate of US+SMI (84.62%, 76.47%, 83.16%) was significantly higher than those of US (78.21%, 64.71%, 75.59%). CONCLUSIONS: The present study supports that SMI is a novel ultrasound technology in revealing micro-vessels in breast lesions. The combined modality of US+SMI presented a better diagnostic performance in making a distinction between PCM and malignant breast carcinomas.