Recent advances in generative models, especially diffusion models, have significantly improved image restoration (IR) performance. However, existing problem-agnostic diffusion model-based image restoration (DMIR) methods face challenges in fully leveraging diffusion priors, resulting in suboptimal performance. In this paper, we address the limitations of current problem-agnostic DMIR methods by analyzing their sampling process and providing effective solutions. We introduce EquS, a DMIR method that imposes equivariant information through dual sampling trajectories. To further boost EquS, we propose the Timestep-Aware Schedule (TAS) and introduce EquS+. TAS prioritizes deterministic steps to enhance certainty and sampling efficiency. Extensive experiments on benchmarks demonstrate that our method is compatible with previous problem-agnostic DMIR methods and significantly boosts their performance without increasing computational costs. Our code is available in https://github.com/FouierL/EquS.

Magnetic Resonance Imaging (MRI), including diffusion MRI (dMRI), serves as a ``microscope'' for anatomical structures and routinely mitigates the influence of low signal-to-noise ratio scans by compromising temporal or spatial resolution. However, these compromises fail to meet clinical demands for both efficiency and precision. Consequently, denoising is a vital preprocessing step, particularly for dMRI, where clean data is unavailable. In this paper, we introduce Di-Fusion, a fully self-supervised denoising method that leverages the latter diffusion steps and an adaptive sampling process. Unlike previous approaches, our single-stage framework achieves efficient and stable training without extra noise model training and offers adaptive and controllable results in the sampling process. Our thorough experiments on real and simulated data demonstrate that Di-Fusion achieves state-of-the-art performance in microstructure modeling, tractography tracking, and other downstream tasks. Code is available at https://github.com/FouierL/Di-Fusion.

Aging is the largest risk factor of major human diseases. Long noncoding RNAs (lncRNAs) as the key regulatory elements have shown a strong impact on multiple biological processes as well as human disease mechanisms. However, the roles of lncRNAs in aging/healthy aging processes remain largely unknown. Centenarians are good models for healthy aging studies due to avoiding major chronic diseases and disabilities. To illustrate their ubiquitous nature in the genome and the 'secrets' of healthy aging regulation from the perspective of lncRNAs, peripheral blood samples from two regions consisting 76 centenarians (CENs), 54 centenarian-children (F1) and 41 spouses of centenarian-children (F1SP) were collected for deep RNA-seq. We identified 11 CEN-specific lncRNAs that is particularly expressed in longevous individuals. By kmers clustering, hundreds of human lncRNAs show similarities with CEN-specific lncRNAs, especially with ENST00000521663 and ENST00000444998. Using F1SP as normal elder controls (age: 59.9 ± 6.6 years), eight lncRNAs that are differentially expressed in longevous elders (CEN group, age: 102.2 ± 2.4 years) were identified as candidate aging/health aging-related lncRNAs (car-lncs). We found that the expression of eight car-lncs in human diploid fibroblasts displayed dynamic changes during cell passage and/or H2O2/rapamycin treatment; of which, overexpression either of THBS1-IT1 and THBS1-AS1, two lncRNAs that highly expressed in CENs, can remarkably decrease p16, p21 and the activity of senescent related β-galactosidase, suggesting that THBS1-IT1 and THBS1-AS1 can inhibit cellular senescence. We provided the first comprehensive analysis of lncRNA expression in longevous populations, and our results hinted that dysregulated lncRNAs in CENs are potential protective factors in healthy aging process.

In radar high-resolution range profile (HRRP)-based target recognition, one of the most challenging tasks is the noncooperative target recognition with imbalanced training data set. This letter presents a novel recognition framework to deal with this problem. The framework is composed of two steps: first, the t-distributed stochastic neighbor embedding (t-SNE) and synthetic sampling are utilized for data preprocessing to provide a well segmented and balanced HRRP data set; second, a discriminant deep belief network (DDBN) is proposed to recognize HRRP data. Compared with the conventional recognition models, the proposed framework not only makes better use of data set inherent structure among HRRP samples for segmentation, but also utilizes high-level features for recognition. Moreover, the DDBN shares latent information of HRRP data globally, which can enhance the ability of modeling the aspect sectors with few HRRP data. The experiments illustrate the meaning of the t-SNE, and validate the effectiveness of the proposed recognition framework with imbalanced HRRP data.