Unpaired Artistic Portrait Style Transfer via Asymmetric Double-Stream GAN.

With the development of image style transfer technologies, portrait style transfer has attracted growing attention in this research community. In this article, we present an asymmetric double-stream generative adversarial network (ADS-GAN) to solve the problems that caused by cartoonization and other style transfer techniques when they are applied to portrait photos, such as facial deformation, contours missing, and stiff lines. By observing the characteristics between source and target images, we propose an edge contour retention (ECR) regularized loss to constrain the local and global contours of generated portrait images to avoid the portrait deformation. In addition, a content-style feature fusion module is introduced for further learning of the target image style, which uses a style attention mechanism to integrate features and embeds style features into content features of portrait photos according to the attention weights. Finally, a guided filter is introduced in content encoder to smooth the textures and specific details of source image, thereby eliminating its negative impact on style transfer. We conducted overall unified optimization training on all components and got an ADS-GAN for unpaired artistic portrait style transfer. Qualitative comparisons and quantitative analyses demonstrate that the proposed method generates superior results than benchmark work in preserving the overall structure and contours of portrait; ablation and parameter study demonstrate the effectiveness of each component in our framework.

Regional Upstroke Tracking for Transit Time Detection to Improve the Ultrasound-Based Local PWV Estimation in Carotid Arteries.

Pulse wave velocity (PWV) is the most important index for quantifying the elasticity of an artery. The accurate estimation of the local PWV is of great relevance to the early diagnosis and effective prevention of arterial stiffness. In ultrasonic transit time-based local PWV estimation, the locations of time fiduciary point (TFP) in the upstrokes of the propagating pulse waves (PWs) are inconsistent because of the reflected waves and ultrasonic noise. In this study, a regional upstroke tracking (RUT) approach that involved identifying the most similar TFP-centered region in the upstrokes is proposed to detect the time delay for improving the local PWV estimation. Five RUT algorithms with different tracking points are assessed via simulation and clinical experiments. To quantitatively evaluate the RUT algorithms, the normalized root-mean-squared errors and standard deviations of the estimated PWVs are calculated using an ultrasound simulation model. The reproducibility of the five RUT algorithms based on 30 human subjects is also evaluated using the Bland-Altman analysis and coefficient of variation (CV). The obtained results show that the RUT algorithms with only three tracking points provide greater accuracy, precision, and reproducibility for the local PWV estimation than the TFP methods. Compared with the TFP methods, the RUT algorithms reduce the mean errors from 12.23% ± 3.10% to 7.13% ± 2.31%, as well as the CVs from 21.76% to 13.39%. In conclusion, the proposed RUT algorithms are superior to the TFP methods for local carotid PWV estimation.

Ultrasound simulation model incorporating incident and reflected wave propagations along a common carotid artery.

An ultrasound simulation model incorporating incident and reflected wave propagations is proposed to provide a realistic data source for validation of transit time (TT)-based local pulse wave velocity (PWV) estimation algorithms. First, the incident wave (IW) and reflected wave (RW) at a certain position over a common carotid artery (CCA) are estimated. Then, the propagating pulse waves (PWs) along the CCA are modelled with the synthesizations of the estimated IWs and RWs, whose occurrences are delayed in opposite sequences according to a preset PWV. In ultrasound simulation, a geometric model of a CCA is built, and the dynamic scatterer models are constructed by moving the scatterer positions according to the synthesized PWs. The RF signals are generated using Field II. To characterize the PW propagations of different arterial stiffnesses consistent with clinical ones in the model, 30 healthy subjects from young, middle-aged, and elderly groups are recruited for extractions of IWs and RWs. To quantitatively verify the effectiveness of the simulation model, the normalized root-mean-squared errors (NRMSEs) are used to compare the estimated and preset PWs, time delays (TDs), and PWVs. Results show that for the three age groups, the estimated PWs, TDs, and PWVs conform to the preset ones with the mean NRMSEs of 0.92%, 18.47%, and 8.55%, respectively. Moreover, the model can characterize the effect of the wave reflection on the local PW propagation as its clinical manifestation. Therefore, the proposed model can be effective as a data source for the validation of TT-based local PWV estimation algorithms, particularly the effects of RWs on the estimation performance.