With the revolution of generative AI, video-related tasks have been widely studied. However, current state-of-the-art video models still lag behind image models in visual quality and user control over generated content. In this paper, we introduce TokenWarping, a novel framework for temporally coherent video translation. Existing diffusion-based video editing approaches rely solely on key and value patches in self-attention to ensure temporal consistency, often sacrificing the preservation of local and structural regions. Critically, these methods overlook the significance of the query patches in achieving accurate feature aggregation and temporal coherence. In contrast, TokenWarping leverages complementary token priors by constructing temporal correlations across different frames. Our method begins by extracting optical flows from source videos. During the denoising process of the diffusion model, these optical flows are used to warp the previous frame's query, key, and value patches, aligning them with the current frame's patches. By directly warping the query patches, we enhance feature aggregation in self-attention, while warping the key and value patches ensures temporal consistency across frames. This token warping imposes explicit constraints on the self-attention layer outputs, effectively ensuring temporally coherent translation. Our framework does not require any additional training or fine-tuning and can be seamlessly integrated with existing text-to-image editing methods. We conduct extensive experiments on various video translation tasks, demonstrating that TokenWarping surpasses state-of-the-art methods both qualitatively and quantitatively. Video demonstrations are available in supplementary materials.
The global incidence of rotator-cuff injuries demands mechanically robust and bioresorbable patches to address the high failure rates of surgical repair. Here, we present a digitally fabricated, warp-knitted silk patch designed to meet this need. Through systematic modulation of key parameters (guide-bar configuration, needle pitch, and gauge), we engineered 17 distinct scaffolds from extra-coarse, degummed silk. Among them, the 4 × 1 tricot-stitch architecture (sample #14) emerged as the optimal candidate, exhibiting an balanced combination of high porosity (70.2 ± 0.1 %), high strength (Longitudinal tensile strength: 274 ± 8 N; Transverse tensile strength: 634 ± 33 N; Longitudinal tear strength: 270 ± 52 N; Transverse tear strength: 125 ± 14 N; Burst strength: 1 554 ± 33 N; Suture-pullout strength: > 46 N.), and controlled biodegradability (retaining 93 ± 3 % mass after 42 days). Critically, these properties not only exceed established mechanical benchmarks for tendon repair but are maintained under wet conditions simulating the in vivo environment. The patch further demonstrated good hemocompatibility (hemolysis rate 1.2 ± 0.1 %) and supported robust cell adhesion, spreading, and proliferation. Collectively, these data demonstrate that digital warp-knitting of coarse silk yarns enables single-step fabrication of lightweight, highly porous, and mechanically anisotropic patches that combine long-term strength retention with favorable biocompatibility-offering a promising off-the-shelf solution for tendon reconstruction. STATEMENT OF SIGNIFICANCE: Recurrent tears after rotator cuff repair remain a significant clinical challenge, often due to inadequate mechanical strength and poor tissue integration of existing patches. We address this by digitally warp-knitting a bioresorbable silk patch that uniquely combines high tensile and burst strength, exceeding native tendon requirements, with a high-porosity architecture conducive to cell infiltration. This patch provides durable mechanical support in a wet physiological environment while degrading controllably. It represents a clinically promising, off-the-shelf solution to enhance repair outcomes, bridging the critical gap between robust mechanical performance and effective biological integration for tendon reconstruction.
Restoring weight is a primary goal during anorexia nervosa (AN) treatment. Previous studies linked different weight gain profiles to treatment outcomes, but there is currently no consensus on profile shapes and numbers. We argue that heterogeneity stems from temporal distortions ("warping") in weight gain, and that similar weight improvements can stretch over different time periods. We thus favor a novel non-parametric solution that accounts for warping to identify weight trajectories. Time series clustering with dynamic time warping (DTW) was used to identify weight change trajectories among N = 518 patients with AN during inpatient treatment. Within-person body-mass-index gain (∆ BMI) served as our primary dependent variable to identify clusters. We characterized clusters based on admission psychopathology scores, and analyzed associations of cluster affiliation with changes in clinical outcomes between admission and discharge using linear and logistic models. We identified four distinct clusters, with n = 76 patients showing initial weight gain (Cluster 1), n = 329 showing continuous weight gain (Cluster 2), n = 70 showing initial weight loss and recovery (Cluster 3), and n = 43 showing weight loss (Cluster 4). The four clusters differed in terms of admission BMI, psychopathology scores, and days spent in treatment, and cluster assignment predicted treatment outcomes. Using one of the largest hitherto examined samples for weight gain profile analysis, the novel DTW-based approach provided an overall more elaborated set of outcome-predictive profiles compared to previous studies, which could help inform individualized treatment strategies and allocate therapeutic resources efficiently.
In this study, biosurfactant-producing bacterial isolates were screened and isolated from a hydrocarbon-rich automobile workshop, marine water, agarwood, and ayurvedic industrial waste. The efficient bacterial isolate Pseudomonas aeruginosa WARP_W1 reduced surface tension to 35.21 mN/m and emulsified 64.32% of olive oil. The biosurfactant production was attempted using different oil sources, with coconut oil producing 899.69 mg/L biosurfactant, which is 35.8% more than olive oil. Logistic kinetic models accurately predicted microbial growth rate (R2 > 0.975) and biosurfactant production rate at 0.348h-1 and 0.201h-1, respectively. These data suggest that coconut oil could be a suitable substrate for biosurfactant. The physicochemical properties were also found to be efficient, with a low critical micelle concentration of 110.60 mg/L and a lowered surface tension of 26.99 mN/m in coconut oil. FTIR and NMR spectroscopy confirmed the glycolipid in the produced biosurfactant by showing rhamnolipid-like structural features. P. aeruginosa WARP_W1 is ideal for large-scale biosurfactant production because of its versatile utilization of carbon sources. This study provides the growth and product kinetic information on the substrate-specific production of biosurfactants.
Warp-knitted spacer fabrics (WKSFs) possess a three-dimensional porous architecture that makes them promising for impact protection and airdrop buffering, yet their lack of intrinsic conductivity and limited cyclic stability restrict intelligent monitoring applications. Here, a structure-function synergistic strategy is proposed by integrating WKSF with carbon-nanotube-modified shear-stiffening gel (cSSG) to construct a conductive, impact-adaptive composite. As a benefit from strain-rate-dependent stiffening and hierarchical energy dissipation, the WKSF-cSSG composite exhibits enhanced impact resistance while forming a stable three-dimensional conductive network. After cyclic preconditioning to suppress the Mullins effect, the composite delivers stable sensing outputs over 3200 cycles with a response time of 18 ms. Under drop-hammer impact, the electrical response shows rapid synchronization with mechanical dynamics, enabling quantitative discrimination of impact intensities. Furthermore, an intelligent airdrop buffering prototype integrating a nine-channel sensing array and deep-learning-assisted classification achieves accurate recognition of five landing postures, demonstrating a material-to-system solution for intelligent protection applications.
Sizing is a critical operation in woven fabric production, as it enhances weaving efficiency by improving warp yarn performance. Conventional sizing agents include maize starch, polyvinyl alcohol (PVA), and commercial carboxymethyl cellulose (CMC). In this study, a low-cost and biodegradable carboxymethyl cellulose derived from wheat straw (CMCws) was investigated as an alternative sizing agent for cotton open-end yarns with a count of Nm 12.2. The high degree of substitution (DS = 1.23) of CMCws indicates extensive carboxymethylation, which enhances the polymer's hydrophilicity and solubility in water. This, in turn, contributes to a higher apparent viscosity (η = 903.03 cP at 300 s-1), reflecting stronger molecular chain interactions and better film-forming ability. CMCws was applied using a high-pressure squeezing technique, and its effect on yarn performance was evaluated in terms of tensile properties, film characteristics, and yarn surface morphology. The results showed that CMCws provided a tenacity gain of 28.57%, a hairiness reduction of 54.34%, and an abrasion resistance gain of 37.14%. These values fall within acceptable industrial ranges and are comparable to those obtained using conventional sizing agents. Furthermore, the optimized CMCws formulation, containing plasticizer and lubricant additives, exhibited good desizing efficiency, with effective removal achieved in hot water. The findings indicate that wheat-straw-derived CMCws is a viable, sustainable alternative to traditional sizing agents for woven fabric production.
Dynamic Time Warping (DTW) is an emerging analytic technique that offers a flexible approach to modeling symptom dynamics in psychological and psychiatric research. Unlike traditional network models, which often rely on linear associations, DTW aligns symptom trajectories even when changes unfold at slightly different speeds or time intervals. This tutorial offers a brief introduction into DTW and demonstrates how to apply DTW to panel or time series data. We illustrate the workflow using clinical case data from patients with eating disorders, to capture temporal patterns that cannot be detected with conventional network analysis techniques, as these require more intensive time-series data. Key advantages include its applicability to non-stationary data, flexibility in handling irregular time intervals, and reduced reliance on frequent assessments, which patients often cannot maintain due to the burden. We also discuss some of the limitations such as noise, scaling decisions and lack of Granger causality associations. Finally, we outline directions for future research. By expanding the methodological toolkit available for studying therapy processes, DTW holds promise for advancing both research and clinical practice in personalized mental health care.
Pentraxins, which constitute a family of evolutionarily conserved pattern recognition molecules, are categorized into short and long branches. The long pentraxin 3 (PTX3) is a key member of the long pentraxin subfamily, while the C-reactive protein and serum amyloid P represent the short pentraxins. All pentraxins share a highly conserved C-terminal motif, an 8-amino acid sequence known as the pentraxin signature. PTX3 can be produced by a wide range of cell types, including immune cells such as dendritic cells, monocytes, and macrophages, as well as various non-immune cells, underscoring its pleiotropic roles in multiple pathophysiological processes. These include inflammation, infection, tissue repair, female fertility, and cancer. Although PTX3 engages commonly recognized signaling pathways, such as TNF-α, NF-κB, FGF, and PI3K/AKT, it can exert paradoxical effects in different cellular contexts, either promoting or inhibiting the proliferation, migration, invasion, and metastasis of cancer cells. This review provides a comprehensive overview of the multifaceted roles of PTX3 in various cancers, while also summarizing its functions in other physiological or pathological contexts. Furthermore, we critically examine the challenges and translational opportunities of PTX3, aiming to inform future research directions and therapeutic strategies for cancer management.
Arthropods, the most diverse phylum on Earth, are hosts to a plethora of bacterial parasites that secrete various effectors of unknown function during infection. The most prevalent of these is the intracellular bacterium Wolbachia pipientis. The microbe infects between 40% and 60% of insect species, where it induces a variety of fitness effects ranging from nutritional supplementation to reproductive manipulations and, in some hosts, limiting virus replication. Understanding the molecular basis of Wolbachia infection and Wolbachia-induced phenotypes is critical to the use of Wolbachia in vector control. Wolbachia ankyrin repeat proteins (WARPs) represent a highly dynamic and diverse part of the Wolbachia pangenome and remain thus far largely uncharacterized. Here, we perform molecular and genetic screens to identify interactions between Wolbachia wMel WARPs and their target host proteins in Drosophila melanogaster. Our results identify strong interactions of two Wolbachia proteins, WARP434 and WARP754, with multiple host targets. Heterologous expression of these two WARPs is extremely toxic in Drosophila tissues, and the toxicity is dependent on the ankyrin repeat domain of each WARP. We use coimmunoprecipitation (coIP) and mass spectrometry to identify native targets of the WARPs, and importantly, knockdown of host targets alleviates toxicity, confirming WARP/target interactions. Antibodies targeting both WARPs show expression by Wolbachia during infection of Drosophila cells, and expression of WARP754 in adult flies increases Wolbachia titer. Understanding how Wolbachia manipulates its host biology and which host pathways it targets during infection will help us define how the most prevalent intracellular bacterial parasite on Earth interacts with its insect hosts at the molecular level. Our screen is an important step toward that goal.IMPORTANCEMolecular interactions drive co-evolutionary arms races between hosts and pathogens. These interactions shape the structure and function of both host and parasite proteins, enabling immunity or virulence during infection. Understanding the molecular details that unfold during these events illustrates not only how hosts and parasites co-evolve at the molecular level but also may help characterize the function of poorly understood proteins. The most prevalent intracellular infection on earth is Wolbachia pipientis, with between 40% and 60% of insects harboring the bacterial symbiont. Understanding how Wolbachia infects host cells and the molecular tools it uses to alter cell biology is critical to the use of the microbe in vector control. Here, we identify Wolbachia proteins used by the symbiont to interface with specific host proteins. Understanding the molecular mechanisms underlying this host-microbe interaction will shed light on how an important symbiont, used in the control of vector populations and disease transmission, uses Wolbachia ankyrin repeat proteins (WARPs) to interact with host targets and how targeting this host protein contributes to infection.
We compared the capabilities of quantitatively assessed paired inspiratory-expiratory area-detector computed tomography (ADCT) for pulmonary functional loss and disease severity evaluations between upright and supine ADCT in matched progressive pulmonary fibrosis (PPF) patients. This retrospective cohort consisted of age-, sex-, and underlying disease-matched patients with PPF who underwent paired inspiratory-expiratory CT on upright ADCT (n = 40) and supine ADCT (n = 40), pulmonary function tests, and disease severity assessment. Based on CT data, the absolute values of the logarithm of the Jacobian determinant and warp-field magnitude of the whole lung and all lobes were calculated. Stepwise regression analyses were performed. On supine ADCT, both indices of the left lower lobe (LLL) were the first and only steps for pulmonary function test results and CT-assessed disease severity (absolute value of the logarithm of the Jacobian determinant: 0.139 ≤ r2 ≤ 0.175, 0.007 ≤ p ≤ 0.018; absolute value of the warp-field magnitude: 0.371 ≤ r2 ≤ 0.447, p < 0.001). However, on upright ADCT, both indices indicated that LLL was the first step and the right lower lobe was the second step for pulmonary function test results and CT-assessed disease severity (0.503 ≤ r2 ≤ 0.674, p < 0.001 or 0.000 < p ≤ 0.006 and 0.474 ≤ r2 ≤ 0.652, 0.002 ≤ p ≤ 0.045, respectively). Upright ADCT has equal to or better potential than supine ADCT for detecting pulmonary functional loss and evaluating disease severity when paired inspiratory-expiratory ADCT is applied in PPF patients. Upright ADCT has superior potential to supine ADCT for pulmonary functional loss and disease severity evaluations when paired inspiratory-expiratory ADCT is performed in patients with progressive pulmonary fibrosis (PPF). Matched progressive pulmonary fibrosis patients compared functional loss and disease severity evaluations between inspiratory-expiratory upright and supine area-detector CT. Clinical parameters demonstrated better correlations with upright than with supine inspiratory-expiratory area-detector CT. Warp-field magnitude showed better correlations with disease severities than the logarithm of the Jacobian determinant on each area-detector CT.
In recent years, nanoparticles and plant-based materials have gained increasing significance in the functional finishing of textile materials. Among natural fibers, cotton is the most widely utilized textile fiber due to its excellent moisture management, superior softness, high absorbency, breathability, compatibility, and biodegradability, making it highly suitable for a broad range of textile and apparel applications. In this study, cotton fabric was successfully functionalized using biosynthesized zinc oxide nanoparticles and Senna didymobotrya leaf extract. ZnCl₂ served as a precursor, while the plant extract acted as both a reducing and functionalizing agent. The dip-dry-cure method was employed with optimized concentrations using Box-Behnken design. Characterization scanning electron microscopy (SEM), X-ray diffraction (XRD), (Fourier Transform Infrared Spectroscopy (FTIR), ultraviolet-visible spectroscopy (UV-vis) confirmed nanoparticle formation with an average size of 20.3 nm. The mechanical results with tensile strength of (warp: 312-318, weft: 289-295 N/mm2) and elongation at break (warp: 22.90-23.40%, weft: 16.29-16.95%), slight stiffness increase (warp: 1.8-1.9 cm, weft: 1.7-1.8 cm), there is a reduced air permeability (16.9-11.8 cm3/cm2/s), with a tear strength (533.36-554.7 g force). The functional properties of the treated cotton fabric have an antibacterial efficiency reached 99.99% (gram-negative bacteria) and 99.5% (gram-positive bacteria), with a ultraviolet radiation protection factor (UPF) of 112.6, indicating strong potential for medical and UV-protective applications.
This study systematically investigates the structural characteristics and antioxidant activities of pectic polysaccharides extracted and purified from Agastache rugosa (Fisch. & C.A.Mey.) Kuntze. Through sequential purification involving ion-exchange and gel permeation chromatography, two homogeneous pectin fractions-WARP-A2b (17.5 kDa) and WARP-A3b (51.5 kDa)-were obtained. Their structural domains, including homogalacturonan, rhamnogalacturonan I, and rhamnogalacturonan II, were characterised using FT-IR, NMR, Congo red binding, circular dichroism, and SEM. Monosaccharide composition analysis revealed both fractions to be rich in galacturonic acid, rhamnose, galactose, and arabinose. WARP-A3b exhibited stronger antioxidant capacity in scavenging ABTS, DPPH, and hydroxyl radicals, which may be attributed to its higher galacturonic acid content and distinct rhamnogalacturonan-I domain organisation. Enzymatic hydrolysis and de-esterification experiments further elucidated the contribution of specific structural domains to antioxidant performance. These results offer new insights into the structure-activity relationships of A. rugosa pectins and support their potential as natural antioxidants in pharmaceuticals.
In this study, an assembly technique based on phytic acid-modified chitosan (PA-CS) and silicone-containing waterborne polyurethane (SiWPU) was developed to construct of a multifunctional hemp fabric integrating flame retardancy, hydrophobicity, and antibacterial properties. A stable multilayer structure was constructed via hydrogen bonding interactions between the phosphate groups of PA-CS and the urethane groups (-NHCOO-) of SiWPU. Research shows that PA-CS significantly enhances the fabric's thermal stability and decomposes at high temperatures to form a dense carbon layer, raising the limiting oxygen index (LOI) of the hemp fabric to 35.1%, achieving a UL-94 V-0 rating. Compared with the original fabric, the PHRR, HRC, and THR of the PA-CS/SiWPU coated hemp fabric were reduced by 80.7%, 80.7%, and 65.8%, respectively. Simultaneously, the silicon component in SiWPU migrates to the fabric surface, resulting in a water contact angle of 133.10°, demonstrating excellent hydrophobicity and antifouling performance. Furthermore, the coating exhibits remarkable antibacterial activity against S. aureus and E. coli. Following the coating treatment, the breaking strength of the fabric increased by 41.6% (warp) and 40.7% (weft), while the elongation at break improved by 28.6% (warp) and 27.9% (weft). Concurrently, the bending rigidity decreased from 4.93 cN/mm to 3.20 cN/mm, accompanied by a reduction in both static and dynamic friction factors. After 10 washing cycles, the fabric exhibited breaking strength retention rates of 94.2% (warp) and 89.1% (weft). This study opens a new avenue for the high-value-added and multifunctional utilization of hemp fabrics, demonstrating great application potential in fields including home, biomedical, and smart textiles.
To achieve near-normal temperature formation of starch paste with satisfactory sizing properties for near-normal temperature sizing application of cotton warps and easy desizing, esterified-sulfonated corn starch (ESCS) samples with degrees of substitution (DSes) of 0.021-0.072 were prepared using dry processes of esterification and sulfonation to evaluate their sizing properties and desizability. It was found that, with the rise in the DSes, the solubility, water dispersibility, adhesion to cotton fibers for the ESCS, and its film elongation and endurance increased, reaching the highest values at DSes = 0.072, which were significantly higher than those of acid-treated starch (ATS). The properties of ESCS (DSes = 0.072) showed no significant variation across temperatures from 95 °C to 40 °C, attributed to only slight variations in solubility. ESCS (DSes = 0.072) could form a paste suitable for immediate use at ~40 °C, with excellent sizing properties for cotton warp sizing at 40 °C to effectively overcome the issues of high-temperature sizing, and easy desizing from cotton warps using a simplified desizing (no chemicals and boiling water used). This study provides valuable methods for the near-normal temperature sizing of starch sizes for cotton warps and easy desizing.
Perception-based Deep Reinforcement Learning (DRL) controllers demonstrate impressive performance on challenging terrains. However, existing controllers still face core limitations, struggling to achieve both terrain generality and platform transferability, and are constrained by high computational overhead and sensitivity to sensor noise. To address these challenges fundamentally, we propose a generalized control framework: Mastering a Generalized Contrastive Depth Model (MGDP). We leverage NVIDIA Warp to enable efficient parallel computation of depth images, thereby mitigating the inherent high computational cost. MGDP extracts low-dimensional terrain feature representations from multi-modal inputs (depth images and height maps) and integrates an explicit depth map denoising mechanism. This process not only facilitates effective decoupling of perception from dynamics but also significantly reduces the memory. Furthermore, we design terrain-adaptive reward functions that modulate penalty strengths according to terrain characteristics, enabling the policy to acquire complex locomotion skills (e.g., climbing, jumping, crawling, squeezing) in a single training stage without relying on distillation. Experimental results demonstrate that MGDP not only endows the policy with superior cross-terrain generalization capability but also enables fast and efficient fine-tuning across diverse quadruped robot morphologies via its pre-trained, dynamics-decoupled perception model. This vigorously advances the development of unified, efficient, and generalized frameworks for quadrupedal locomotion control.
The rapid development of COVID-19 vaccines as a result of Operation Warp Speed represented an extraordinary triumph of scientific innovation, with multiple vaccine platforms demonstrating remarkable efficacy in preventing severe disease, hospitalization, and death-ultimately saving millions of lives. The United States government, pharmaceutical companies, research institutions, and health agencies collaborated at an unprecedented scale to develop, test, and distribute vaccines, showcasing what coordinated medical innovation can accomplish. Vaccination programs successfully prevented catastrophic outcomes and enabled people to return to normal life during the crisis. COVID-19 vaccines continue to provide critical protection for vulnerable populations today, and the mRNA platforms developed have opened new possibilities for treating cancers and other diseases. Yet despite rigorous regulatory oversight and extensive clinical trial data, these vaccines have faced substantial misinformation campaigns that spread false claims about their safety and design. To address these misrepresentations and resultant public concerns, this document draws on rigorous scientific evidence to comprehensively examine the misconceptions surrounding mRNA technology, explaining how these vaccines work, their proven safety record, and their demonstrated benefits.
The COVID-19 pandemic created an urgent need to develop preventive vaccines to blunt the impact of the greatest global health crisis in a century. Two vaccines, both employing mRNA technology, were developed from bench to bedside in under one year - a milestone considered nearly impossible. This paper examines the evolving safety profile of mRNA-1273, from development under Operation Warp Speed through Emergency Use Authorization, and subsequent deployment at nearly unprecedented scale. Vaccine safety was characterized during clinical development and refined through well-established safety monitoring systems (e.g. Vaccine Adverse Event Reporting System [VAERS]), as well as newly introduced systems such as V-Safe. Key safety findings, such as myocarditis, are reviewed as well as safety findings in special populations. The complementary contributions of public, private, and academic sectors highlight how collaboration and rigorous monitoring supported timely and comprehensive safety assessment of a new vaccine in the extraordinary setting of a global pandemic.
Respiratory-induced motion complicates accurate irradiation of thoraco-abdominal tumors during radiotherapy, as treatment-system latency entails target-location uncertainties. This work addresses frame forecasting in dynamic chest and liver MRI to compensate for such delays. We investigate RNNs trained with online learning algorithms, enabling adaptation to changing respiratory patterns via on-the-fly parameter updates, and transformers, increasingly common in time-series forecasting for their ability to capture long-term dependencies. Experiments were conducted using twelve sagittal thoracic and upper-abdominal cine-MRI sequences from ETH Zürich and Otto-von-Guericke University Magdeburg (OvGU); the OvGU data exhibited higher motion variability, noise, and lower contrast. PCA decomposes the Lucas-Kanade optical-flow field into static deformation modes and low-dimensional, time-dependent weights. We compare various methods for forecasting these weights: linear filters, population and sequence-specific encoder-only transformers, and RNNs trained with real-time recurrent learning (RTRL), unbiased online recurrent optimization, decoupled neural interfaces, and sparse one-step approximation (SnAp-1). Predicted displacements were used to warp the reference frame and generate future images. Prediction accuracy decreased with the horizon h. Linear regression performed best at short horizons (1.3 mm geometrical error at h=0.32s, ETH Zürich dataset), while RTRL and SnAp-1 outperformed the other algorithms at medium-to-long horizons, with geometrical errors below 1.4 mm and 2.8 mm on the sequences from ETH Zürich and OvGU, respectively. The sequence-specific transformer was competitive for low-to-medium horizons, but transformers remained overall limited by data scarcity and domain shift between datasets. Predicted frames visually resembled the ground truth, with notable errors occurring near the diaphragm at end-inspiration and regions affected by out-of-plane motion.
Segmentation of cardiac structures is essential for cardiac function evaluation using cine magnetic resonance imaging (MRI). Deep learning can be used to segment cardiac structures in cine cardiac MRI with high accuracy, but this approach requires fully annotated datasets for training, which are difficult to obtain. Semi-supervised segmentation methods provide a way to alleviate the burden of manual labeling by using labeled and unlabeled data for training. However, these methods generally provide suboptimal segmentation accuracies. To develop a semi-supervised method that utilizes relatively small training datasets and under-annotations for improved cine cardiac MRI segmentation. The proposed approach consists of deformable registration, fully and weakly supervised segmentation, and a temporal attention perceiver (TAP). The registration module was trained to warp labeled frames to generate pseudo labels for unlabeled frames. The warped labeled images were used to train the fully supervised segmentation network. The unlabeled images and the pseudo label were used to train the weakly supervised segmentation model, and the segmentation prediction was compared with the input pseudo label as an auxiliary loss to the registration module. The TAP module was employed to generate optimized features for the warped labeled and the original unlabeled images both paired with the original labeled image. Consistency between the resulting features was enforced to refine cross-instance feature alignment to facilitate the registration. One hundred, twenty, and ten subjects from the Automatic Cardiac Diagnosis Challenge (ACDC) and seventy-five, thirty, and fifteen cases from the Multi-Vendor & Multi-Disease (M&Ms) Cardiac Image Segmentation Challenge were used for training, each with random end-systolic (ES)/end-diastolic (ED) frames labeled. The optimized models were used to segment the remaining 50 ACDC and 50 M&Ms subjects. The proposed approach was compared with several commonly used semi-supervised segmentation methods in terms of Dice-similarity-coefficients (DSC), average-symmetric-surface-distance (ASSD), and Hausdorff-distance (HD) for left (LV) and right (RV) ventricular cavity and myocardium (Myo). A Unet trained on the same subjects each with both frames labeled was used as an upper bound (Unet_UB). Using 100 ACDC training subjects, our approach yielded DSC = 0.910 ± $\pm$ 0.063, ASSD = 1.37 ± $\pm$ 0.63 mm, and HD = 6.38 ± $\pm$ 2.99 mm for RV, DSC = 0.894 ± $\pm$ 0.024, ASSD = 1.20 ± $\pm$ 1.12 mm, and HD = 4.67 ± $\pm$ 3.22 mm for Myo, and DSC = 0.934 ± $\pm$ 0.056, ASSD = 1.25 ± $\pm$ 1.63 mm, and HD = 3.97 ± $\pm$ 5.76 mm for LV. A bidirectional copy-paste (BCP) method performed the best among the comparative methods and generated DSC = 0.902 ± $\pm$ 0.060, ASSD = 1.45 ± $\pm$ 0.60 mm, and HD = 7.50 ± $\pm$ 3.20 mm for RV, DSC = 0.885 ± $\pm$ 0.030, ASSD = 1.28 ± $\pm$ 0.80 mm, and HD = 5.80 ± $\pm$ 2.80 mm for Myo, and DSC = 0.920 ± $\pm$ 0.068, ASSD = 1.15 ± $\pm$ 0.40 mm, and HD = 4.20 ± $\pm$ 3.30 mm for LV. For Unet_UB, these were 0.905 ± $\pm$ 0.068, 1.48 ± $\pm$ 0.61 mm, and 6.35 ± $\pm$ 2.85 mm for RV, 0.895 ± $\pm$ 0.030, 1.05 ± $\pm$ 0.45 mm, and 4.40 ± $\pm$ 3.09 mm for Myo, and 0.941 ± $\pm$ 0.044, 1.02 ± $\pm$ 0.34 mm, and 3.17 ± $\pm$ 1.63 mm for LV. Similar trends were observed when using 75 M&Ms training subjects. For all the experiments, our approach outperformed BCP in general and yielded segmentation accuracies comparable to Unet_UB. The proposed approach outperformed several commonly used semi-supervised segmentation methods and yielded segmentation accuracies on par with fully supervised Unet using various relatively small datasets and under annotations for training.