A vehicular antenna with diverse radiation patterns is proposed for both the dedicated short range communications (DSRC) and the fifth-generation (5 G) millimeter-wave (MMW) communications. The proposed antenna mainly comprised a 2×2 MMW subarray, an 8-element DSRC subarray and a corresponding 5-port feeding network. The radiating patches of the MMW subarray are excited by two pairs of coaxial-to-substrate integrated waveguide (SIW) transition (CST) feeding networks that are comprised of two 3-dB couplers, two 90 $^{\circ }$ -phase shifters, and four power dividers (PDs), which provide in-phase and differential-phase output signals to achieve single-beam and conical-beam radiation patterns. On the other hand, the 8 monopoles of the DSRC subarray arranged around the aforementioned MMW radiator are excited by a coaxial-to-SIW-to- coaxial transition (CSCT) feeding network that includes a 1-to-4 SIW PD and four 1-to-2 microstrip PDs to form an omni-directional radiation and achieve a large frequency ratio (LFR). Measured results indicate that wide bandwidths of 15.44% (24.14-28.18 GHz), 16.09% (24.29-28.54 GHz), 15.01% (23.85-27.72 GHz), and 14.29% (24.05-27.75 GHz) with corresponding high peak gains of 10.97 $\pm$ 1.03 dBi, 11.71 $\pm$ 1.32 dBi, 10.84 $\pm$ 1.00 dBi, and 11.29 $\pm$ 1.28 dBi are obtained by the four MMW subarray ports, while a bandwidth of 11.40% (5.46-6.12 GHz) with an azimuth-plane gain of 0.64 dBi and a peak gain of 4.29 $\pm$ 0.19 dBi for the DSRC subarray port are also obtained. Besides, low cross-polarization (X-pol) levels of -24 dB and -27 dB are achieved for the two subarrays, respectively.
Ultrasound (US) image segmentation embraced its significant improvement in deep learning era. However, the lack of sharp boundaries in US images still remains an inherent challenge for segmentation. Previous methods often resort to global context, multi-scale cues or auxiliary guidance to estimate the boundaries. It is hard for these methods to approach pixel-level learning for fine-grained boundary generating. In this paper, we propose a novel and effective framework to improve boundary estimation in US images. Our work has three highlights. First, we propose to formulate the boundary estimation as a rendering task, which can recognize ambiguous points (pixels/voxels) and calibrate the boundary prediction via enriched feature representation learning. Second, we introduce point-wise contrastive learning to enhance the similarity of points from the same class and contrastively decrease the similarity of points from different classes. Boundary ambiguities are therefore further addressed. Third, both rendering and contrastive learning tasks contribute to consistent improvement while reducing network parameters. As a proof-of-concept, we performed validation experiments on a challenging dataset of 86 ovarian US volumes. Results show that our proposed method outperforms state-of-the-art methods and has the potential to be used in clinical practice.