Entropy is a quantitative measure of signal uncertainty and has been widely applied to ultrasound tissue characterization. Ultrasound assessment of hepatic steatosis typically involves a backscattered statistical analysis of signals based on information entropy. Deep learning extracts features for classification without any physical assumptions or considerations in acoustics. In this study, we assessed clinical values of information entropy and deep learning in the grading of hepatic steatosis. A total of 205 participants underwent ultrasound examinations. The image raw data were used for Shannon entropy imaging and for training and testing by the pretrained VGG-16 model, which has been employed for medical data analysis. The entropy imaging and VGG-16 model predictions were compared with histological examinations. The diagnostic performances in grading hepatic steatosis were evaluated using receiver operating characteristic (ROC) curve analysis and the DeLong test. The areas under the ROC curves when using the VGG-16 model to grade mild, moderate, and severe hepatic steatosis were 0.71, 0.75, and 0.88, respectively; those for entropy imaging were 0.68, 0.85, and 0.9, respectively. Ultrasound entropy, which varies with fatty infiltration in the liver, outperformed VGG-16 in identifying participants with moderate or severe hepatic steatosis (p < 0.05). The results indicated that physics-based information entropy for backscattering statistics analysis can be recommended for ultrasound diagnosis of hepatic steatosis, providing not only improved performance in grading but also clinical interpretations of hepatic steatosis.

High-intensity focused ultrasound (HifU) is a well-accepted tool for noninvasive thermal therapy. to control the quality of HifU treatment, the focal spot generated in tissues must be localized. Ultrasound imaging can monitor heated regions; in particular, the change in backscattered energy (cBe) allows parametric imaging to visualize thermal information in the tissue. conventional cBe imaging constructed in the spatial domain may be easily affected by noises when the HIFU focal spot is visualized. this study proposes frequency-domain cBe imaging to improve noise tolerance and image contrast in HifU focal spot monitoring. phantom experiments were performed in a temperaturecontrolled environment. HIFU of 2.12 MHz was applied to the phantoms, during which a clinical scanner equipped with a 3-MHz convex array transducer was used to collect raw image data consisting of backscattered signals for B-mode, spatial-, and frequency-domain cBe imaging. concurrently, temperature changes were measured at the focal spot using a thermocouple for comparison with CBE values by calculating the correlation coefficient r. to further analyze cBe image contrast levels, a contrast factor was introduced, and an independent t-test was performed to calculate the probability value p. experimental results showed that frequency-domain cBe imaging performed well in thermal distribution visualization, enabling quantitative detection of temperature changes. the cBe value calculated in the frequency domain also correlated strongly with that obtained using the conventional spatial-domain approach (r = 0.97). In particular, compared with the image obtained through the conventional method, the contrast of the cBe image obtained using the method based on frequencydomain analysis increased by 2.5-fold (4 dB; p < 0.05). Frequency-domain computations may constitute a new strategy when ultrasound cBe imaging is used to localize the focal spot in HifU treatment planning.

Duchenne muscular dystrophy (DMD) results in loss of ambulation and premature death. Ultrasound provides real-time, safe, and cost-effective routine examinations. Deep learning allows the automatic generation of useful features for classification. This study utilized deep learning of ultrasound imaging for classifying patients with DMD based on their ambulatory function. A total of 85 individuals (including ambulatory and nonambulatory subjects) underwent ultrasound examinations of the gastrocnemius for deep learning of image data using LeNet, AlexNet, VGG-16, VGG-16TL, VGG-19, and VGG-19TL models (the notation TL indicates fine-tuning pretrained models). Gradient-weighted class activation mapping (Grad-CAM) was used to visualize features recognized by the models. The classification performance was evaluated using the confusion matrix and receiver operating characteristic (ROC) curve analysis. The results show that each deep learning model endows muscle ultrasound imaging with the ability to enable DMD evaluations. The Grad-CAMs indicated that boundary visibility, muscular texture clarity, and posterior shadowing are relevant sonographic features recognized by the models for evaluating ambulatory function. Of the proposed models, VGG-19 provided satisfying classification performance (the area under the ROC curve: 0.98; accuracy: 94.18%) and feature recognition in terms of physical characteristics. Deep learning of muscle ultrasound is a potential strategy for DMD characterization.