The rapid growth of large models' size has far outpaced that of computing resources. To bridge this gap, encouraged by the parsimonious relationship between genotype and phenotype in the brain's growth and development, we propose the so-called Hyper-Compression that turns the model compression into the issue of parameter representation via a hyperfunction. Specifically, it is known that the trajectory of some low-dimensional dynamic systems can fill the high-dimensional space eventually. Thus, Hyper-Compression, using these dynamic systems as the hyperfunctions, represents the parameters of the target network by their corresponding composition number or trajectory length. This suggests a novel mechanism for model compression, substantially different from the existing pruning, quantization, distillation, and decomposition. Along this direction, we methodologically identify a suitable dynamic system with the irrational winding as the hyperfunction and theoretically derive its associated error bound. Next, guided by our theoretical insights, we propose several engineering twists to make the Hyper-Compression pragmatic and effective. Lastly, systematic and comprehensive experiments on NLP models such as LLaMA and Qwen series and vision models confirm that Hyper-Compression enjoys the following PNAS merits: 1) Preferable compression ratio; 2) No post-hoc retraining; 3) Affordable inference time; and 4) Short compression time. It compresses LLaMA2-7B in an hour and achieves close-to-int4-quantization performance, without retraining and with a performance drop of less than 1%.

Limited-angle computed tomography (LACT) improves temporal resolution and reduces radiation dose, but suffers from severe artifacts due to missing projections. Clinical workflows record abundant patient- and acquisition-level metadata, yet such information remains underutilized in image reconstruction. To tackle the ill-posed LACT inverse problem, we propose a metadata-guided two-stage diffusion framework that leverages structured clinical contexts as semantic priors for robust reconstruction. In Stage-I, we learn a metadata-to-anatomy generative prior by conditioning a transformer-based diffusion model on clinical metadata (acquisition parameters, patient demographics, and diagnostic impressions), and sampling a coarse anatomical estimate from Gaussian noise. In Stage-II, a second conditional diffusion model performs coarse-to-fine refinement, using the Stage-I estimate as an image prior while re-injecting the same metadata to recover full-resolution anatomy. To preserve anatomical fidelity and suppress hallucinations, projection-domain data consistency is enforced periodically after denoising update via an ADMM-based solver. Experiments on the public multimodal CTRATE dataset demonstrate that the proposed framework outperforms iterative, CNN-based, and diffusion-based baselines, with the greatest gains under severe truncation, e.g., up to 5.23%/11.21% higher SSIM/PSNR than the strongest metadata-free diffusion competitor at 90°. On real clinical cardiac CT, it yields coronary artery calcium scores closer to full-view references, indicating improved clinical utility. Furthermore, the proposed method is generalized to out-of-distribution angular ranges and projection geometries, and ablation results confirm complementary contributions from different metadata types under limited-angle conditions. Our results highlight clinical metadata as actionable semantic priors to synergize with physics-informed constraints to improve both reconstruction fidelity and clinical quantification in LACT.

Cone-beam Computed Tomography (CBCT) provides real-time three-dimensional (3D) imaging support for intraoperative navigation. However, high-attenuation metal implants introduce severe metal artifacts in reconstructed CBCT images. These artifacts compromise image quality and therefore may affect diagnostic accuracy. Current CBCT metal artifact reduction (MAR) algorithms overlook the complementary information available across CBCT views, leading to inaccurate projection-domain interpolation and secondary artifacts in the reconstructed images. To tackle these challenges, we propose a novel Unsupervised Projection-domain Multiview Constraint Learning Network (UPMCL-Net), which directly learns from metal-affected data for CBCT MAR without ground truths. In addition, a transformer-based MultiView Consistency Module (MVCM) is constructed to interpolate the projection-domain metal region for cross-view consistency. Finally, a Hybrid Feature Attention Module (HFAM) is designed to adaptively fuse interview and intraview features. Comprehensive experiments conducted on real clinical datasets confirm the performance of UPMCL-Net, showcasing its potential as an efficient, accurate, and reliable approach for CBCT MAR in clinical intraoperative interventions.

Contrast-enhanced MRI enhances pathological visualization but often necessitates Pre-Contrast images for accurate quantitative analysis and comparative assessment. However, Pre-Contrast images are frequently unavailable due to time, cost, or safety constraints, or they may suffer from degradation, making alignment challenging. This limitation hinders clinical diagnostics and the performance of tools requiring combined image types. To address this challenge, we propose a novel staged, physics-grounded learning framework with a hyperintensity prior to synthesize Pre-Contrast images directly from Post-Contrast MRIs. The proposed method can generate high-quality Pre-Contrast images, thus, enabling comprehensive diagnostics while reducing the need for additional imaging sessions, costs, and patient risks. To the best of our knowledge, this is the first Pre-Contrast synthesis model capable of generating images that may be interchangeably used with standard-of-care Pre-Contrast images. Extensive evaluations across multiple datasets, sites, anatomies, and downstream tasks demonstrate the model’s robustness and clinical applicability, positioning it as a valuable tool for contrast-enhanced MRI workflows.

Cardiac computed tomography (CT) has emerged as a major imaging modality for the diagnosis and monitoring of cardiovascular diseases. High temporal resolution is essential to ensure diagnostic accuracy. Limited-angle data acquisition can reduce scan time and improve temporal resolution, but typically leads to severe image degradation and motivates for improved reconstruction techniques. In this paper, we propose a novel physics-informed score-based diffusion model (PSDM) for limited-angle reconstruction of cardiac CT. At the sampling time, we combine a data prior from a diffusion model and a model prior obtained via an iterative algorithm and Fourier fusion to further enhance the image quality. Specifically, our approach integrates the primal-dual hybrid gradient (PDHG) algorithm with score-based diffusion models, thereby enabling us to reconstruct high-quality cardiac CT images from limited-angle data. The numerical simulations and real data experiments confirm the effectiveness of our proposed approach.

A neural network (NN) with the widely-used ReLU activation has been shown to partition the sample space into many convex polytopes for prediction. However, the parametric way a NN and other machine learning models use to partition the space has imperfections, e.g., the compromised interpretability for complex models, the inflexibility in decision boundary construction due to the generic character of the model, and the risk of being trapped into shortcut solutions. In contrast, although the nonparameterized models can adorably avoid or downplay these issues, they are usually insufficiently powerful either due to over-simplification or the failure to accommodate the manifold structures of data. In this context, we first propose a new type of machine learning models referred to as Manifoldron that directly derives decision boundaries from data and partitions the space via manifold structure discovery. Then, we systematically analyze the key characteristics of the Manifoldron such as manifold characterization capability and its link to NNs. The experimental results on four synthetic examples, 20 public benchmark datasets, and one real-world application demonstrate that the proposed Manifoldron performs competitively compared to the mainstream machine learning models. We have shared our code in https://github.com/wdayang/Manifoldron for free download and evaluation.

Low-dose computed tomography (LDCT) offers reduced X-ray radiation exposure but at the cost of compromised image quality, characterized by increased noise and artifacts. Recently, transformer models emerged as a promising avenue to enhance LDCT image quality. However, the success of such models relies on a large amount of paired noisy and clean images, which are often scarce in clinical settings. In computer vision and natural language processing, masked autoencoders (MAE) have been recognized as a powerful self-pretraining method for transformers, due to their exceptional capability to extract representative features. However, the original pretraining and fine-tuning design fails to work in low-level vision tasks like denoising. In response to this challenge, we redesign the classical encoder-decoder learning model and facilitate a simple yet effective streamlined low-level vision MAE, referred to as LoMAE, tailored to address the LDCT denoising problem. Moreover, we introduce an MAE-GradCAM method to shed light on the latent learning mechanisms of the MAE/LoMAE. Additionally, we explore the LoMAE's robustness and generability across a variety of noise levels. Experimental findings show that the proposed LoMAE enhances the denoising capabilities of the transformer and substantially reduce their dependency on high-quality, ground-truth data. It also demonstrates remarkable robustness and generalizability over a spectrum of noise levels. In summary, the proposed LoMAE provides promising solutions to the major issues in LDCT including interpretability, ground truth data dependency, and model robustness/generalizability.

In Cognitive Radio (CR) networks, it is common that the spectrum holes are too narrow to support high-speed communications. Discontinuous Orthogonal Frequency Division Multiplexing (DOFDM) is a good way for a secondary user to access several spectrum fragments simultaneously with one Radio Front (RF). In this paper, a novel Channel Characteristic Aware Spectrum Aggregation (CCASA) algorithm which uses DOFDM to aggregation spectrum fragments with only one radio front is proposed in order to increase the overall throughput of a CR network. By combining Adaptive Modulation and Coding (AMC) and spectrum aggregation, the good subcarriers are assigned to the specific secondary users in CCASA algorithm thus achieving a better channel efficiency. Different bandwidth requirement and aggregation limitation of secondary users are both considered in this algorithm while maintaining a fairly low computational complexity. The simulation results show that CCASA achieves a bigger total throughput than existing aggregation algorithms.