Reference-based MRI reconstruction uses fully-sampled images to facilitate the reconstruction of under-sampled data, enhancing imaging efficiency and quality. However, most existing reference-based approaches require strict spatial and modality correspondence between the reference and target images, a condition rarely satisfied in routine clinical practice and one that limits their broader deployment. To overcome this limitation, we propose URG-MRI, a generative-prior framework that distils high-quality anatomical priors from unpaired reference images, removing the need for the reference and target to share subject, contrast, or spatial registration. The method follows a two-stage training strategy: a vector-quantized generative model first learns a discrete codebook of informative features from fully-sampled images; a feature-consistency loss then guides the retrieval of relevant priors during under-sampled reconstruction, while spatial and distribution alignment modules mitigate residual misalignments between retrieved priors and target features. The framework is designed as a plug-in module compatible with diverse reconstruction backbones. On the IXI and fastMRI datasets, integrating it into six representative networks (U-Net, KIKI-Net, DuDoRNet, HUMUS-Net, FPSFormer, and E2E-VarNet) yields PSNR improvements across 4 × /8 × Random and Equispaced under-sampling, with representative gains of +1.01 dB for U-Net on IXI (4 × Random) and up to +1.63 dB for KIKI-Net on fastMRI (8 × Equispaced). Paired t-test and Wilcoxon signed-rank tests further confirm that the improvements are statistically significant (p < .001) in nearly all configurations. Our code will be available at: https://github.com/chenxm12394/URG-MRI.
The NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Vaginal Cancer outline the recommended diagnostic workup, staging considerations, and treatment options for this rare malignancy. Because vaginal cancer is uncommon and shares many biologic and clinical characteristics with cervical cancer, several management recommendations, particularly systemic therapy, are extrapolated from evidence and practices established for cervical cancer. This guideline excerpt summarizes key components of management, including diagnostic evaluation and workup, principles of staging, pathology considerations, radiation therapy principles, and primary treatment recommendations for both early-stage and advanced disease. It also details approaches to manage relapses, including locoregional recurrence and distant metastases, and provides an in-depth overview of systemic therapy recommendations for vaginal cancer. J Natl Compr Canc Netw 2026;24(3):101-126 doi:10.6004/jnccn.2026.0011
Students in high school and college are expected to take on greater responsibility for regulating their own learning. As they navigate course demands and explore potential career paths, students are making decisions about what to attend to, how to engage, how much effort to invest, and how long to persist. Motivation and its regulation is key to this process (Sansone et al., 2019). We integrate insights about the importance of task value from situated expectancy-value theory (Eccles Wigfield, 2020) and research on utility-value interventions (UVIs) that target task value with insights from the four-phase model of interest development (Renninger Hidi, 2022) to identify ways to support students as they manage motivational challenges. Value and interest are often closely related, and, when considered separately, each positively predicts student effort. Rather than considering them in isolation, we consider how they may work together over time. UVIs can help students think about and articulate the usefulness of course topics, which has been found to enhance perceived task value and motivate engagement with course content (Harackiewicz Priniski, 2018; Wigfield Eccles, 2020). UVIs may be especially helpful when students have little interest in learning the content, but this effect may become more nuanced as students develop interest. We suggest a framework for integrating these perspectives, identify self-regulation challenges, and discuss possible ways to support this self-regulatory process. We also highlight future directions for research that emerge from this integration.
Thermal boundary conductance (TBC) between silicon and diamond influences heat removal in next-generation electronics where Si and diamond serve as substrates or channels. In this paper, we report simulations of Si/diamond interfaces using machine-learning interatomic potential (MLIP)-driven molecular dynamics. MLIPs trained on ab initio Si/diamond heterostructures faithfully reproduce interfacial Si-C bonding and the phonon dispersions of bulk Si and diamond, whereas the classical Tersoff potential fails to capture these accurately. After quantum correction, the room-temperature TBC of a bare Si/diamond interface is 130 MW m-2 K-1, in close agreement with experiment (approximately 140 MW m-2 K-1), and far below predictions from Tersoff molecular dynamics (310 MW m-2 K-1) or a first-principles phonon-dispersion diffuse mismatch model (205 MW m-2 K-1). Adding a 1-nm-thick SiNx interlayer approximately doubles the TBC to 275 MW m-2 K-1, and replacing SiNx with amorphous carbon further increases it to 350 MW m-2 K-1. The density of states analysis reveals that the interlayer supplies a large portion of intermediate phonon modes that bridge Si and diamond. Finite-element device simulations indicate that these TBC improvements measurably reduce peak chip temperature. These results highlight interfacial-layer engineering as a potential route to improved thermal management in wide-band-gap electronics.
Point defects are fundamental imperfections in crystalline materials spontaneously formable from thermal excitations and mediate a wide range of phenomena from phase transformations to semiconductor technologies. Despite their ubiquity and importance, it was not until 1926 that point defects were first postulated by Yakov Frenkel. This issue celebrates the 100-year anniversary of this conceptual breakthrough and the technological impacts from metallurgy, to energy, to quantum technologies it has enabled. This collection of six articles presents a broad (but not exhaustive) snapshot of current research on defects in materials—a mini-preview of the next 100 years.
