Head movement poses a significant challenge in brain positron emission tomography (PET) imaging, resulting in image artifacts and tracer uptake quantification inaccuracies. Effective head motion estimation and correction are crucial for precise quantitative image analysis and accurate diagnosis of neurological disorders. Hardware-based motion tracking (HMT) has limited applicability in real-world clinical practice. To overcome this limitation, we propose a deep-learning head motion correction approach with cross-attention (DL-HMC++) to predict rigid head motion from one-second 3D PET raw data. DL-HMC++ is trained in a supervised manner by leveraging existing dynamic PET scans with gold-standard motion measurements from external HMT. We evaluate DL-HMC++ on two PET scanners (HRRT and mCT) and four radiotracers (18F-FDG,18F-FPEB,11C-UCB-J, and11C-LSN3172176) to demonstrate the effectiveness and generalization of the approach in large cohort PET studies. Quantitative and qualitative results demonstrate that DL-HMC++ consistently outperforms state-of-the-art data-driven motion estimation methods, producing motion-free images with clear delineation of brain structures and reduced motion artifacts that are indistinguishable from gold-standard HMT. Brain region of interest standard uptake value analysis exhibits average difference ratios between DL-HMC++ and gold-standard HMT to be 1.2±0.5% for HRRT and 0.5±0.2% for mCT. DL-HMC++ demonstrates the potential for data-driven PET head motion correction to remove the burden of HMT, making motion correction accessible to clinical populations beyond research settings. The code is available at https://github.com/maxxxxxxcai/DL-HMC-TMI.

Unsupervised Domain Adaptation (UDA), which aligns the labeled source distribution to the unlabeled target distribution, has shown remarkable achievement in the medical image segmentation task. Previous UDA methods unilaterally consider the global distribution alignment through explicit category-based loss while good separation and discrimination of class are insufficiently explored, resulting in the sub-aligned distribution across domains. In this paper, we propose cross-prototype contrastive learning method (CPCL) for UDA segmentation through class centroid alignment. Specifically, to reduce the intra-class distance and increase the inter-class distance, we first introduce prototype-feature contrastive learning to align the pixel-level features and the same-class global prototype across domains. Secondly, we further present prototype-prototype contrastive learning to align the same class prototypes between the source domain and target domain for compact category centroid and better global domain distribution alignment. Extensive experiments on two public cardiac datasets demonstrate that the proposed CPCL achieves superior domain adaptation performance as compared with the state-of-the-art.

A two-stage design procedure combining the strength of a fixed-time sliding mode control and a high-gain compensator for deadzone nonlinearity is proposed for a multimotor driving system. A novel practical fixed-time convergent controller is designed for the perturbed system, which improves the applicability of the proposed method. The concept of multisurface sliding mode is used to cope with the load tracking problem and can guarantee fixed-time convergence, which is regardless of initial states of the system. The convergence time can be known as a priori and a satisfactory dynamic performance can be obtained. Meanwhile, the fixed-time convergent synchronization controller is designed to guarantee the synchronization of driving motors. Then, a high-gain nonlinearity compensator is designed to reduce performance degradation caused by the deadzone nonlinearity. Its simple form makes it more practical and reliable to use than existing compensation methods. Comparative experimental results demonstrate the efficacy of the developed control scheme.

The wave excitation force estimation and prediction play an important role in improving the performance of causal and noncausal controllers for wave energy converters (WECs). This article proposes a robust adaptive sliding-mode observer (ASMO) to estimate the wave excitation force subject to unknown disturbances and parametric uncertainties for a multimotion multifloat WEC, called M4. Both the convergence time and the estimation error can be explicitly bounded within expected limits by tuning the ASMO parameters, which are essentially beneficial for causal controllers to maintain the control performance. A fixed-time convergent sliding variable is designed to drive the estimation error into a small region within a fixed time. Due to the adaptive law, the overall system is proven to be finite-time stable, which allows explicit formulations of the convergence time and the estimation error. Moreover, based on the wave force estimation by the ASMO, an improved auto-regressive (AR) model whose coefficients are updated by online training is developed to predict the wave excitation force. The prediction errors can also be explicitly estimated to achieve guaranteed control performance for the noncausal controller requiring future excitation force. From the comparison based on a realistic sea wave gathered from Cornwall, U.K., it can be found that compared with the conventional Kalman filter, the ASMO achieves a smaller steady-state estimation error and has satisfactory robustness performance against 30% model mismatch.

In the time division duplex (TDD) communication system, the sounding reference signal (SRS) is used for uplink channel estimation and we can eventually obtain the downlink channel depending on the channel reciprocity. However, the limited SRS makes it impossible to estimate the channels of all users simultaneously if there are numerous users in a cell. This can degrade the performance of multi-user (MU) grouping and scheduling. In this paper, we present a user grouping scheduling scheme for TDD massive MIMO system under the SRS limited condition. The algorithm alliances several consecutive time and frequency domain resources to form a resource block group (RBG) set. The users are grouped and assigned to one RBG in a RBG set for channel estimation and scheduling, according to the approximate correlation through the channel information obtained by users in RBG. Under the SRS limited situation, simulation results show that the proposed scheme can increase the throughput of the system compared to the schemes without considering this impact.

Identifying the transmission status as line-of-sight (LOS) and non-line-of-sight (NLOS) is of importance for 3D Massive Multiple-Input Multiple-Output (MIMO) systems, which is one of the core technology to improve 5G New Radio (NR) capacity and spectral efficiency. If the identification could be as accurate as possible, the positioning systems and adaptive radio systems like cognitive radios can increase their performances significantly. This paper presents an improved algorithm for LOS/NLOS identification in 3D Massive MIMO systems. By fully considering the characteristics of the 3D MIMO channel, we formulate the identification problem as a binary hypothesis test by exploiting a statistic model based on time-space-frequency channel correlation. Compared to a previous study based on channel correlation, our simulation results show that the performance of the novel method is over 8.7% improved, and the error rate is as low as 3.23%. Besides, the effect of the number of antennas and taps in time domain on the performance of the improved method is discussed.

Channel estimation is one of the most important parts in three‐dimensional multiple‐input multiple‐output (3D MIMO) systems. The characteristics of non‐line‐of‐sight (NLOS) channel and line‐of‐sight (LOS) channel are different in 3D MIMO systems. If the same channel estimation scheme is used in LOS case as NLOS, the performance of estimation will be bad. In outdoor propagation environment, 3D MIMO channels between closely located antennas share the same delay support in temporal domain. With those prior knowledge, in this study, a new channel estimation scheme is proposed. The proposed scheme can be divided into two processes. First, it is needed to identify the received sounding reference signal whether is LOS or NLOS propagation. Then, different enhanced DFT‐based channel estimation schemes are proposed separately according to the identification results. Simulation results verify the proposed algorithm outperforms traditional discrete Fourier transform (DFT)‐based channel estimation. At signal‐to‐noise ratio of 20 dB, the proposed algorithm has 17.7 and 35.7% improvement in NLOS case and LOS case separately in terms of normalised mean squared error compared with traditional DFT‐based channel estimation scheme, and is achieved with additional liner complexity.

In time-division duplex (TDD) massive MIMO systems, multi-user (MU) grouping and scheduling is a key technology to improve spectral efficiency. However, in practice limited sounding reference signals (SRS) restrict the channel information acquired by base station (BS), which degrades the performance of MU grouping. To solve this problem, a joint grouping and scheduling scheme based on channel state information reference signal (CSI-RS) is proposed. Our scheme includes two processes. In the first process, we propose a metric to measure the channel accuracy of users estimated several periods before with the aid of un-precoded and beamformed CSI-RS. In the second process, a joint grouping and scheduling algorithm is implemented to distribute low-correlation users with acceptable channel accuracy into one group. Both low-speed and medium-speed circumstances are considered in our simulation using 3D MIMO channel model. In the cases of limited SRS resources, compared to the schemes that do not consider this impact, results show our proposed scheme can enhance the performance of users with low throughputs. Further, a better system sum rate can be reached at the same time.

In this paper, an approximation-free funnel feedback controller is proposed for a class of nonlinear servomechanisms to achieve prescribed tracking error performance. An improved funnel function is proposed to guarantee the transient and asymptotic behavior of the tracking error within a given funnel boundary. The proposed funnel function removes the imposed assumption used in conventional funnel controls (e.g., systems with relative degree one or two) and avoids the potential singularity problem in prescribed performance controls. Moreover, an extended state observer (ESO) is used to address the effect of unknown dynamics in the control system (e.g., friction and disturbances), where the ESO parameters can be easily designed based on the control system bandwidth. The stability of the proposed control system with ESO and funnel function is analyzed via the Lyapunov theory. Comparative simulations and experimental results are conducted based on a practical turntable servomechanisms to validate the efficacy of the proposed method.