Central and peripheral analgesic active components of triterpenoid saponins from Stauntonia chinensis and their action mechanism.

Triterpenoid saponins from Stauntonia chinensis have been proven to be a potential candidate for inflammatory pain relief. Our pharmacological studies confirmed that the analgesic role of triterpenoid saponins from S. chinensis occurred via a particular increase in the inhibitory synaptic response in the cortex at resting state and the modulation of the capsaicin receptor. However, its analgesic active components and whether its analgesic mechanism are limited to this are not clear. In order to further determine its active components and analgesic mechanism, we used the patch clamp technique to screen the chemical components that can increase inhibitory synaptic response and antagonize transient receptor potential vanilloid 1, and then used in vivo animal experiments to evaluate the analgesic effect of the selected chemical components. Finally, we used the patch clamp technique and molecular biology technology to study the analgesic mechanism of the selected chemical components. The results showed that triterpenoid saponins from S. chinensis could enhance the inhibitory synaptic effect and antagonize the transient receptor potential vanilloid 1 through different chemical components, and produce central and peripheral analgesic effects. The above results fully reflect that "traditional Chinese medicine has multi-component, multi-target, and multi-channel synergistic regulation".

ECC-PolypDet: Enhanced CenterNet With Contrastive Learning for Automatic Polyp Detection.

Accurate polyp detection is critical for early colorectal cancer diagnosis. Although remarkable progress has been achieved in recent years, the complex colon environment and concealed polyps with unclear boundaries still pose severe challenges in this area. Existing methods either involve computationally expensive context aggregation or lack prior modeling of polyps, resulting in poor performance in challenging cases. In this paper, we propose the Enhanced CenterNet with Contrastive Learning (ECC-PolypDet), a two-stage training & end-to-end inference framework that leverages images and bounding box annotations to train a general model and fine-tune it based on the inference score to obtain a final robust model. Specifically, we conduct Box-assisted Contrastive Learning (BCL) during training to minimize the intra-class difference and maximize the inter-class difference between foreground polyps and backgrounds, enabling our model to capture concealed polyps. Moreover, to enhance the recognition of small polyps, we design the Semantic Flow-guided Feature Pyramid Network (SFFPN) to aggregate multi-scale features and the Heatmap Propagation (HP) module to boost the model's attention on polyp targets. In the fine-tuning stage, we introduce the IoU-guided Sample Re-weighting (ISR) mechanism to prioritize hard samples by adaptively adjusting the loss weight for each sample during fine-tuning. Extensive experiments on six large-scale colonoscopy datasets demonstrate the superiority of our model compared with previous state-of-the-art detectors.

NG-NAS: Node growth neural architecture search for 3D medical image segmentation.

Neural architecture search (NAS) has been applied to design proper 3D networks for medical image segmentation. In order to reduce the computation cost in NAS, researchers tend to adopt weight sharing mechanism to search architectures in a supernet. However, recent studies state that the searched architecture rankings may not be accurate with weight sharing mechanism because the training situations are inconsistent between the searching and training phases. In addition, some NAS algorithms design inflexible supernets that only search operators in a pre-defined backbone and ignore the importance of network topology, which limits the performance of searched architecture. To avoid weight sharing mechanism which may lead to inaccurate results and to comprehensively search network topology and operators, we propose a novel NAS algorithm called NG-NAS. Following the previous studies, we consider the segmentation network as a U-shape structure composed of a set of nodes. Instead of searching from the supernet with a limited search space, our NG-NAS starts from a simple architecture with only 5 nodes, and greedily grows the best candidate node until meeting the constraint. We design 2 kinds of node generations to form various network topological structures and prepare 4 candidate operators for each node. To efficiently evaluate candidate node generations, we use NAS without training strategies. We evaluate our method on several public 3D medical image segmentation benchmarks and achieve state-of-the-art performance, demonstrating the effectiveness of the searched architecture and our NG-NAS. Concretely, our method achieves an average Dice score of 85.11 on MSD liver, 65.70 on MSD brain, and 87.59 in BTCV, which performs much better than the previous SOTA methods.

Hierarchical Weight Averaging for Deep Neural Networks.

Despite simplicity, stochastic gradient descent (SGD)-like algorithms are successful in training deep neural networks (DNNs). Among various attempts to improve SGD, weight averaging (WA), which averages the weights of multiple models, has recently received much attention in the literature. Broadly, WA falls into two categories: 1) online WA, which averages the weights of multiple models trained in parallel, is designed for reducing the gradient communication overhead of parallel mini-batch SGD and 2) offline WA, which averages the weights of one model at different checkpoints, is typically used to improve the generalization ability of DNNs. Though online and offline WA are similar in form, they are seldom associated with each other. Besides, these methods typically perform either offline parameter averaging or online parameter averaging, but not both. In this work, we first attempt to incorporate online and offline WA into a general training framework termed hierarchical WA (HWA). By leveraging both the online and offline averaging manners, HWA is able to achieve both faster convergence speed and superior generalization performance without any fancy learning rate adjustment. Besides, we also analyze the issues faced by the existing WA methods, and how our HWA addresses them, empirically. Finally, extensive experiments verify that HWA outperforms the state-of-the-art methods significantly.

Uniaxial Strain and Hydrostatic Pressure Engineering of the Hidden Magnetism in La1-xCaxMnO3 (0 ≤ x ≤ 1/2) Thin Films.

Here, using various substrates, we demonstrate that the in-plane uniaxial strain engineering can enhance the Jahn-Teller distortions and promote selective orbital occupancy to induce an emergent antiferromagnetic insulating (AFI) phase at x = 1/3 of La1-xCaxMnO3. Such an AFI phase depends not only on the magnitude of epitaxial strain but also on the symmetry of the substrates. Using the large uniaxial strain imparted by DyScO3(001) substrate, the AFI ground state is achieved in a wide range of doping levels (0 ≤ x ≤ 1/2), leaving an extended AFI phase diagram. Moreover, it is found that hydrostatic pressure can tune the AFI phase back to a hidden ferromagnetic metallic phase, accompanied by the formation of accommodation strain. The coaction of the accommodation strain, uniaxial strain, and hydrostatic pressure produces complex phase competition and evolution, and the result may shed light on phase space control of other functional perovskites with the competing magnetic interactions.

Misfit Relaxation Mechanisms and Domain Ordering in Anisotropically Strained Manganite Thin Films.

The evolution of anisotropic strain in epitaxial Pr0.5Sr0.5MnO3 films grown on (LaAlO3)0.3(SrAl0.5Ta0.5O3)0.7(110) substrates has been characterized by off-specular X-ray reciprocal space mappings on the (130), (310), (222), and (222̅) reflections in the scattering zone containing the [110] axis. We demonstrate that a multistage hierarchical structural evolution (single-domain-like structure, domain ordering, twin domains, and/or periodic structural modulations) occurs as the film thickness increases, and the structural modulation between the two transverse in-plane [11̅0] and [001] directions is quite different due to the monoclinic distortion of the film. We then show the relationship between the distribution of diffraction spots in reciprocal space and their corresponding domain configurations in real space under various thicknesses, which is closely correlated with thickness-dependent magnetic and magnetotransport properties. More importantly, the distribution and annihilation dynamics of the domain ordering are imaged utilizing home-built magnetic force microscope, revealing that the structural domains tilted toward either the [001] or [001̅] direction are arranged along the [11̅1] and [1̅11] crystal orientations. The direct visualization and dynamics of anisotropic-strain-related domain ordering will open a new path toward the control and manipulation of domain engineering in strongly correlated perovskite oxide films.

Uniaxial Strain-Controlled Ground States in Manganite Films.

Strongly correlated perovskite oxides exhibit a plethera of intriguing phenomena and stimulate a great potential for multifunctional device applications. Utilizing tunable uniaxial strain, rather than biaxial or anisotropic strain, delivered from the crystallography of a single crystal substrate to modify the ground state of strongly correlated perovskite oxides has rarely been addressed for phase-space control. Here, we show that the physical properties of La2/3Ca1/3MnO3 (LCMO) films are remarkably different depending on the crystallographic orientations of the orthorhombic NdGaO3 (NGO) substrates. More importantly, the antiferromagnetic charge-ordered insulating (COI) phase induced in the (100) or (001)-oriented LCMO films can be dramatically promoted (or suppressed) by a uniaxial tensile (or compressive) bending stress along the in-plane [010] direction. By contrast, the COI phase is nearly unaffected along the other transverse in-plane directions. Results from scanning transmission electron microscopy reveal that the (100)- or (001)-oriented LCMO films are uniaxially tensile strained along the [010] direction, while the LCMO/NGO(010) and LCMO/NGO(110) films remaining as a bulklike ferromagnetic metallic state exhibit a different strain state. Density functional theory calculations further reveal that the cooperatively increased Jahn-Teller distortion and charge ordering may be indispensible for the inducing and promoting of the COI phase. These findings provide a path to understand the correlation between local and extended structural distortions imparted by coherent epitaxy and the electronic states for quantum phase engineering.

Robust Ferroelectric Properties in (K,Na)NbO3-Based Lead-Free Films via a Self-Assembled Nanocomposite Approach.

(K,Na)NbO3-based lead-free ferroelectric materials are highly desired in modern electronic applications and have long been considered as a strong candidate for replacing (Pb,Zr)TiO3, but most of them are deficient in large remnant polarization and decent thermal stability. Here, a unique lead-free 0.95(K0.49Na0.49Li0.02)(Nb0.8Ta0.2)O3-0.05CaZrO3 with 2 wt % MnO2 addition (KNNLT-CZ-M) ferroelectric film with special nanocomposite structures grown on La0.7Sr0.3MnO3-coated SrTiO3(001) substrate is demonstrated. The KNNLT-CZ-M films display excellent ferroelectricity with a large twice remnant polarization of 64.91 µC/cm2, a superior thermal stability of ferroelectricity from -196 to 300 °C, and a high Curie temperature of 400 °C. These robust performances could be attributed to the densely arranged self-assembled nanocolumns (∼10 nm in diameter) in the films, which can vertically strain the matrix and enhance its b/a ratio. The formation of the nanocolumns critically depends on the CaZrO3 component. Our results may help the design of a new type of lead-free ferroelectric films and promote their potential applications in microelectronic devices.