Ultra-Stable High-Capacity Polythiophene Derivative for Wide-Potential-Window Supercapacitors.

Conducting polymer (CP)-based supercapacitors show great promise for applications in the field of wearable and portable electronics. However, these supercapacitors face persistent challenges, notably low energy density and inadequate stability. In this study, we introduce a polythiophene derivative, designated as poly(EPE), synthesized via the electrochemical polymerization of 8-bis(2,3-dihydrothieno[3,4-b][1,4]dioxin-5-yl)-3,3-dimethyl-3,4-dihydro-2H-thieno[3,4-b][1,4]dioxepine (EPE). The resulting poly(EPE) polymer exhibits an exemplary 3D porous network-like structure, significantly enhancing its capacitance performance. When employed as the electrode material, the symmetric supercapacitor demonstrates an exceptionally high specific capacitance of 1342 F g-1 at a current density of 4.0 A g-1, along with impressive energy and power densities of 119.3 W h kg-1 and 38.83 kW kg-1, respectively. These capacitance values surpass those of previously reported pristine CP-based supercapacitors. Notably, the supercapacitor showcases outstanding stability, maintaining a retention rate of 92.5% even after 50,000 charge-discharge cycles. These findings underscore the substantial potential of poly(EPE) as an electrode material for the advancement of the supercapacitor technology.

Pulmonary blastoma is successfully treated with immunotherapy and targeted therapy.

Pulmonary blastomas (PB) are an extremely rare type of lung cancer. Currently, no standard treatment exists for PB. Immunotherapy with checkpoint inhibitors and anti-angiogenesis treatments has been an effective method for lung cancer; however, studies on PB treatment are lacking. Herein, we present a case report of successful conversion therapy with immunotherapy and targeted therapy for PB. After receiving treatment with a PD-1 inhibitor (penpulimab) and a multi-target tyrosine kinase inhibitor (anlotinib) treatment, the patient showed an impressive response and underwent a successful operation. We also summarized and reviewed literature reports on PubMed from January 1, 2000, to December 31, 2022, using the keyword "pulmonary blastoma", discussing the efficacy and specifics of chemotherapy and radiotherapy. Immunotherapy, in combination with targeted therapy, should be considered a potential therapeutic strategy for PB.

Selenium-SelK-GPX4 axis protects nucleus pulposus cells against mechanical overloading-induced ferroptosis and attenuates senescence of intervertebral disc.

Intervertebral disc degeneration (IVDD) is one of the most prevalent spinal degenerative disorders and imposes places heavy medical and economic burdens on individuals and society. Mechanical overloading applied to the intervertebral disc (IVD) has been widely recognized as an important cause of IVDD. Mechanical overloading-induced chondrocyte ferroptosis was reported, but the potential association between ferroptosis and mechanical overloading remains to be illustrated in nucleus pulposus (NP) cells. In this study, we discovered that excessive mechanical loading induced ferroptosis and endoplasmic reticulum (ER) stress, which were detected by mitochondria and associated markers, by increasing the intracellular free Ca2+ level through the Piezo1 ion channel localized on the plasma membrane and ER membrane in NP cells. Besides, we proposed that intracellular free Ca2+ level elevation and the activation of ER stress are positive feedback processes that promote each other, consistent with the results that the level of ER stress in coccygeal discs of aged Piezo1-CKO mice were significantly lower than that of aged WT mice. Then, we confirmed that selenium supplementation decreased intracellular free Ca2+ level by mitigating ER stress through upregulating Selenoprotein K (SelK) expression. Besides, ferroptosis caused by the impaired production and function of Glutathione peroxidase 4 (GPX4) due to mechanical overloading-induced calcium overload could be improved by selenium supplementation through Se-GPX4 axis and Se-SelK axis in vivo and in vitro, eventually presenting the stabilization of the extracellular matrix (ECM). Our findings reveal the important role of ferroptosis in mechanical overloading-induced IVDD, and selenium supplementation promotes significance to attenuate ferroptosis and thus alleviates IVDD, which might provide insights into potential therapeutic interventions for IVDD.

Cortistatin prevents glucocorticoid-associated osteonecrosis of the femoral head via the GHSR1a/Akt pathway.

Long-term use of glucocorticoids (GCs) is known to be a predominant cause of osteonecrosis of the femoral head (ONFH). Moreover, GCs can mediate apoptosis of various cell types by exaggerating oxidative stress. We have previously found that Cortistatin (CST) antagonizes oxidative stress and improves cell apoptosis in several conditions. In this study, we detected that the CST expression levels were diminished in patients with ONFH compared with femoral neck fracture (FNF). In addition, a GC-induced rat ONFH model was established, which impaired bone quality in the femoral head. Then, administration of CST attenuated these ONFH phenotypes. Furthermore, osteoblast and endothelial cells were cultured and stimulated with dexamethasone (Dex) in the presence or absence of recombinant CST. As a result, Dex induced impaired anabolic metabolism of osteoblasts and suppressed tube formation in endothelial cells, while additional treatment with CST reversed this damage to the cells. Moreover, blocking GHSR1a, a well-accepted receptor of CST, or blocking the AKT signaling pathway largely abolished the protective function of CST in Dex-induced disorder of the cells. Taken together, we indicate that CST has the capability to prevent GC-induced apoptosis and metabolic disorder of osteoblasts in the pathogenesis of ONFH via the GHSR1a/AKT signaling pathway.

Development of Convergent Hybrid Phase Ligation for Efficient and Convenient Total Synthesis of Proteins.

Simple and efficient total synthesis of homogeneous and chemically modified protein samples remains a significant challenge. Here, we report development of a convergent hybrid phase native chemical ligation (CHP-NCL) strategy for facile preparation of proteins. In this strategy, proteins are split into ~100-residue blocks, and each block is assembled on solid support from synthetically accessible peptide fragments before ligated together into full-length protein in solution. With the new method, we increase the yield of CENP-A synthesis by 2.5-fold compared to the previous hybrid phase ligation approach. We further extend the new strategy to the total chemical synthesis of 212-residue linker histone H1.2 in unmodified, phosphorylated, and citrullinated forms, each from eight peptide segments with only one single purification. We demonstrate that fully synthetic H1.2 replicates the binding interactions of linker histones to intact mononucleosomes, as a proxy for the essential function of linker histones in the formation and regulation of higher order chromatin structure.

Combine radiotherapy and immunotherapy in esophageal squamous cell carcinoma.

Immune checkpoint inhibitors(ICIs) have improved the survival of advanced esophageal squamous cell carcinoma (ESCC) patients. Radiotherapy is one of the common therapies to treat esophageal cancer. However, whether combination radiation therapy can increase the efficacy of immunotherapy is still up for debate. Radiotherapy combined with immunotherapy has proven to be a reliable and effective treatment for tumors, and it can work in combination with immunotherapy to achieve better anti-tumor effects. This review aims to discuss the efficacy and safety of combining radiotherapy and immunotherapy to treat ESCC by stages as well as the optimum radiotherapy dose and target volume, with a summary of clinical trials in ESCC.

Selenomethionine against titanium particle-induced osteolysis by regulating the ROS-dependent NLRP3 inflammasome activation via the ß-catenin signaling pathway.

Wear debris-induced osteolysis, especially titanium (Ti) particles-induced osteolysis, is the most common cause of arthroplasty failure with no effective therapy. Previous studies have suggested that inflammation and impaired osteogenesis are associated with Ti particles -induced osteolysis. Selenium (Se) is an essential trace element in the human body, which forms selenomethionine (Se-Met) in nature, and selenoproteins has strong anti-inflammatory and antioxidant stress effects. In this study, the effects of Se-Met on Ti particles-induced osteolysis were observed and the potential mechanism was explored. We found that exogenous Se-Met relieved osteolysis induced by Ti particles in two animal models and MC3T3-E1 cells. We found that the addition of Se-Met effectively inhibited Ti particle-induced inflammation by regulating reactive oxygen species-dependent (ROS-dependent) NOD-like receptor protein 3 (NLRP3) inflammasome activation. These therapeutic effects were abrogated in MC3T3-E1 cells that had received a ß-catenin antagonist, suggesting that Se-Met alleviates inflammatory osteolysis via the ß-catenin signaling pathway. Collectively, these findings indicated that Se-Met may serve as a potential therapeutic agent for treating Ti particle-induced osteolysis.

Metformin attenuates chronic lung allograft dysfunction: evidence in rat models.

BACKGROUND: Chronic lung allograft dysfunction (CLAD) directly causes an abysmal long-term prognosis after lung transplantation (LTx), but effective and safe drugs are not available. Metformin exhibits high therapeutic potential due to its antifibrotic and immunomodulatory effects; however, it is unclear whether metformin exerts a therapeutic effect in CLAD. We sought to investigate the effect of metformin on CLAD based on rat models. METHODS: Allogeneic LTx rats were treated with Cyclosporin A (CsA) in the first week, followed by metformin, CsA, or vehicle treatment. Syngeneic LTx rats received only vehicles. All rats were sacrificed on post-transplant week 4. Pathology of lung graft, spleen, and thymus, extent of lung fibrosis, activity of profibrotic cytokines and signaling pathway, adaptive immunity, and AMPK activity were then studied. RESULTS: Allogeneic recipients without maintenance CsA treatment manifested CLAD pathological characteristics, but these changes were not observed in rats treated with metformin. For the antifibrotic effect, metformin suppressed the fibrosis extent and profibrotic cytokine expression in lung grafts. Regarding immunomodulatory effect, metformin reduced T- and B-cell infiltration in lung grafts, spleen and thymus weights, the T- and B-cell zone areas in the spleen, and the thymic medullary area. In addition, metformin activated AMPK in lung allografts and in α-SMA+ cells and T cells in the lung grafts. CONCLUSIONS: Metformin attenuates CLAD in rat models, which could be attributed to the antifibrotic and immunomodulatory effects. AMPK activation suggests the potential molecular mechanism. Our study provides an experimental rationale for further clinical trials.

Pararespiratory and paradigestive lymph node metastases in esophageal squamous cell carcinoma: predicting survival and refining the N staging system.

BACKGROUND: The site of lymph node metastasis (LNM) may affect the prognosis of patients with esophageal squamous cell carcinoma (ESCC). To investigate the prognoses of pararespiratory and paradigestive LNM and to propose a novel N (nN) staging system that integrates both the LNM site and count. METHODS: This study was a multicenter, large-sample, retrospective cohort study that included ESCC patients with LNM between January 2014 and December 2019 from three Chinese institutes. Patients were set into training (two institutes) and external validation (one institute) cohorts. The primary outcomes were survival differences in LNM site and the development of novel nodal staging system. The overall survival (OS) of patients with pararespiratory LNM only (Group A), paradigestive LNM only (Group B), and both sites (Group C) was evaluated by Kaplan-Meier. Cox proportional hazards models were used to identify the independent prognostic factors. An nN staging system considering both the LNM site and count was developed and evaluated by the area under the receiver operating characteristic curve (AUC). RESULTS: In total, 1313 patients were included and split into training (n = 1033) and external validation (n = 280) cohorts. There were 342 (26.0%), 568 (43.3%) and 403 (30.7%) patients in groups A, B and C, respectively. The OS of patients with pararespiratory and patients with paradigestive LNM presented significant differences in the training and validation cohorts (P < 0.050). In the training cohort, LNM site was an independent prognostic factor (hazard ratio: 1.58, 95% confidence intervals: 1.41-1.77, P < 0.001). The nN staging definition: nN1 (1-2 positive pararespiratory/paradigestive LNs), nN2 (3-6 pararespiratory LNs or 1 pararespiratory with 1paradigestive LN), nN3 (3-6 LNs with ≥ 1 paradigestive LN), nN4 (≥ 7 LNs). Subsets of patients with different nN stages showed significant differences in OS (P < 0.050). The prognostic model of the nN staging system presented higher performance in the training and validation cohorts at 3-year OS (AUC, 0.725 and 0.751, respectively) and 5-year OS (AUC, 0.740 and 0.793, respectively) than the current N staging systems. CONCLUSIONS: Compared to pararespiratory LNM, the presence of paradigestive LNM is associated with worse OS. The nN staging system revealed superior prognostic ability than current N staging systems.

Visible trephine-based foraminoplasty in PTED leads to asymmetrical stress changes and instability in the surgical and adjacent segments: a finite element analysis.

This study aimed to construct a multi-segment lumbar finite element model (FEM) of PTED surgery to analyze the changes in stress and ROM after visible trephine-based foraminoplasty. The CT scans of a 35-year-old healthy male were used to develop a multi-segment lumbar FEM with Mimic, Geomagic Studio, Hypermesh and MSC.Patran. Different foraminoplasty was performed on the model, and these were grouped into normal group (A), the ventral resection group (B), the apex resection group (C), the ventral + apex + isthmus resection group (D), and the SAP + isthmus + lateral recess resection group (E). A vertical load of 500N and a torque of 10N·M were applied to the upper surface of the L3 vertebral body to simulate the biomechanical characteristics under the motion of flexion, extension, lateral bending, and rotation. The von Mises stress maps of the intervertebral f, vertebral body, facet joints, and the ROM of the L3-S1 intervertebral disk were calculated and analyzed. The changes of peak stress on the vertebral body for each group were not significant in the same motion state. Significant stress differences were observed in the L4/5 intervertebral disks, while no obvious stress changes were observed for the L3/4 and L5/S1 intervertebral disks. The stress of the L3/4 and L5/S1 facet joints decreased after L4/5 foraminoplasty, while the stress of L4/5 facet joints displayed an overall increasing trend. Significant asymmetrical stress changes of bilateral facet joints were observed in all three segments, particularly during bilateral rotation movements. The ROM of L3-S1 gradually increased from Group A to Group E, especially during flexion, left lateral bending, and right rotation, with the highest elevation observed for the L45 ROM. Our FEM indicated that enlarged resection and exposure of the articular surface could lead to significant asymmetrical stress changes in the bilateral facet joints and ROM instability of the surgical and adjacent segments. These findings suggested that unnecessary and excessive resection should be avoided in PTED to reduce the incidence of low back pain and the risk of postsurgical degeneration.

Generation of Perfect Electron Vortex Beam with a Customized Beam Size Independent of Orbital Angular Momentum.

The electron vortex beam (EVB)-carrying quantized orbital angular momentum (OAM) plays an essential role in a series of fundamental research. However, the radius of the transverse intensity profile of a doughnut-shaped EVB strongly depends on the topological charge of the OAM, impeding its wide applications in electron microscopy. Inspired by the perfect vortex in optics, herein, we demonstrate a perfect electron vortex beam (PEVB), which completely unlocks the constraint between the beam size and the beam's OAM. We design nanoscale holograms to generate PEVBs carrying different quanta of OAM but exhibiting almost the same beam size. Furthermore, we show that the beam size of the PEVB can be readily controlled by only modifying the design parameters of the hologram. The generation of PEVB with a customized beam size independent of the OAM can promote various in situ applications of free electrons carrying OAM in electron microscopy.

Learning View-Based Graph Convolutional Network for Multi-View 3D Shape Analysis.

View-based approach that recognizes 3D shape through its projected 2D images has achieved state-of-the-art results for 3D shape recognition. The major challenges are how to aggregate multi-view features and deal with 3D shapes in arbitrary poses. We propose two versions of a novel view-based Graph Convolutional Network, dubbed view-GCN and view-GCN++, to recognize 3D shape based on graph representation of multiple views. We first construct view-graph with multiple views as graph nodes, then design two graph convolutional networks over the view-graph to hierarchically learn discriminative shape descriptor considering relations of multiple views. Specifically, view-GCN is a hierarchical network based on two pivotal operations, i.e., feature transform based on local positional and non-local graph convolution, and graph coarsening based on a selective view-sampling operation. To deal with rotation sensitivity, we further propose view-GCN++ with local attentional graph convolution operation and rotation robust view-sampling operation for graph coarsening. By these designs, view-GCN++ achieves invariance to transformations under the finite subgroup of rotation group SO(3). Extensive experiments on benchmark datasets (i.e., ModelNet40, ScanObjectNN, RGBD and ShapeNet Core55) show that view-GCN and view-GCN++ achieve state-of-the-art results for 3D shape classification and retrieval tasks under aligned and rotated settings.

Hypoxia-Inducible Factor-1α Protects Against Intervertebral Disc Degeneration Through Antagonizing Mitochondrial Oxidative Stress.

Intervertebral disc degeneration (IVDD) demonstrates a gradually increased incidence and has developed into a major health problem worldwide. The nucleus pulposus is characterized by the hypoxic and avascular environment, in which hypoxia-inducible factor-1α (HIF-1α) has an important role through its participation in extracellular matrix synthesis, energy metabolism, cellular adaptation to stresses and genesis. In this study, the effects of HIF-1α on mouse primary nucleus pulposus cells (MNPCs) exposed to TNF-α were observed, the potential mechanism was explored and a rabbit IVDD model was established to verify the protective role of HIF-1α on IVDD. In vitro results demonstrated that HIF-1α could attenuate the inflammation, apoptosis and mitochondrial dysfunction induced by TNF-α in MNPCs; promote cellular anabolism; and inhibit cellular catabolism. In vivo results demonstrated that after establishment of IVDD model in rabbit, disc height and IVD extracellular matrix were decreased in a time-dependent manner, MRI analysis showed a tendency for decreased T2 values in a time-dependent manner and supplementation of HIF-1α improved histological and imaginative IVDD while downregulation of HIF-1α exacerbated this degeneration. In summary, HIF-1α protected against IVDD, possibly through reducing ROS production in the mitochondria and consequent inhibition of inflammation, metabolism disorders and apoptosis of MNPCs, which provided a potential therapeutic instrument for the treatment of IVDD diseases.

circ-Iqsec1 induces bone marrow-derived mesenchymal stem cell (BMSC) osteogenic differentiation through the miR-187-3p/Satb2 signaling pathway.

BACKGROUND: Bone marrow-derived mesenchymal stem cells (BMSCs) are general progenitor cells of osteoblasts and adipocytes and they are characterized as a fundamental mediator for bone formation. The current research studied the molecular mechanisms underlying circRNA-regulated BMSC osteogenic differentiation. METHODS: Next-generation sequencing (NGS) was employed to study abnormal circRNA and mRNA expression in BMSCs before and after osteogenic differentiation induction. Bioinformatics analysis and luciferase reporting analysis were employed to confirm correlations among miRNA, circRNA, and mRNA. RT-qPCR, ALP staining, and alizarin red staining illustrated the osteogenic differentiation ability of BMSCs. RESULTS: Data showed that circ-Iqsec1 expression increased during BMSC osteogenic differentiation. circ-Iqsec1 downregulation reduced BMSC osteogenic differentiation ability. The present investigation discovered that Satb2 played a role during BMSC osteogenic differentiation. Satb2 downregulation decreased BMSC osteogenic differentiation ability. Bioinformatics and luciferase data showed that miR-187-3p linked circ-Iqsec1 and Satb2. miR-187-3p downregulation or Satb2 overexpression restored the osteogenic differentiation capability of BMSCs post silencing circ-Iqsec1 in in vivo and in vitro experiments. Satb2 upregulation restored osteogenic differentiation capability of BMSCs post miR-187-3p overexpression. CONCLUSION: Taken together, our study found that circ-Iqsec1 induced BMSC osteogenic differentiation through the miR-187-3p/Satb2 signaling pathway.

Preoperative Clinical Characteristics Predict Recurrent Laryngeal Nerve Lymph Node Metastasis and Overall Survival in Esophageal Squamous Cell Carcinoma: A Retrospective Study With External Validation.

Background: Recurrent laryngeal nerve (RLN) lymph node metastasis (LNM) is not rare in patients with esophageal squamous cell carcinoma (ESCC). We aimed to develop and externally validate a preoperative nomogram using clinical characteristics to predict RLN LNM in patients with ESCC and evaluate its prognostic value. Methods: A total of 430 patients with ESCC who underwent esophagectomy with lymphadenectomy of RLN LNs at two centers between May 2015 and June 2019 were reviewed and divided into training (center 1, n = 283) and external validation cohorts (center 2, n = 147). Independent risk factors for RLN LNM were determined by multivariate logistic regression, and a nomogram was developed. The performance of the nomogram was assessed in terms of discrimination, calibration, clinical usefulness, and prognostic value. The nomogram was internally validated by the bootstrap method and externally validated by the external validation cohort. Results: Multivariate analysis indicated that clinical T stage (P <0.001), endoscopic tumor length (P = 0.003), bioptic tumor differentiation (P = 0.004), and preoperative carcinoembryonic antigen level (P = 0.001) were significantly associated with RLN LNM. The nomogram had good discrimination with the area under the curve of 0.770 and 0.832 after internal and external validations. The calibration curves and decision curve analysis confirmed the good calibration and clinical usefulness of this model. High-risk of RLN LNM predicted by the nomogram was associated with worse overall survival in the external validation cohort (P <0.001). Conclusion: A nomogram developed by preoperative clinical characteristics demonstrated a good performance to predict RLN LNM and prognosis for patients with ESCC.

Synthesis of Hollow N,P-Doped Carbon/Co2P2O7 Nanotubular Crystals as an Effective Electrocatalyst for the Oxygen Reduction Reaction.

Herein, N,P-rich carbon/carbon/Co2P2O7 hollow nanotubes with a multilayered wall structure were successfully fabricated for the ORR electrocatalyst. The hollow tube structure catalysts were obtained by carbonizing Co2P2O7/C coated with the phytate-doped PANI. The Co2P2O7/C was obtained by phosphorylating a basic cobalt carbonate with phytic acid (PA). Onset and positive half-wave potentials were measured at 0.90 and 0.84 V, respectively, with a diffusion-limited current density of 4.58 mA/cm2. Effect of the thickness of polyaniline (PANI) in the electrocatalyst precursor was also investigated. The specific surface area as well as the content of graphitic N altered as the time of PANI polymerization increased, resulting in remarkably different catalytic activities. This study of hollow nanotube catalysts exhibits efficient noble-metal-free oxygen reduction reaction electrocatalysts for other chemical systems, which will provide abundant electrochemical active centers and sufficient energy.

Learning Polynomial-Based Separable Convolution for 3D Point Cloud Analysis.

Shape classification and segmentation of point cloud data are two of the most demanding tasks in photogrammetry and remote sensing applications, which aim to recognize object categories or point labels. Point convolution is an essential operation when designing a network on point clouds for these tasks, which helps to explore 3D local points for feature learning. In this paper, we propose a novel point convolution (PSConv) using separable weights learned with polynomials for 3D point cloud analysis. Specifically, we generalize the traditional convolution defined on the regular data to a 3D point cloud by learning the point convolution kernels based on the polynomials of transformed local point coordinates. We further propose a separable assumption on the convolution kernels to reduce the parameter size and computational cost for our point convolution. Using this novel point convolution, a hierarchical network (PSNet) defined on the point cloud is proposed for 3D shape analysis tasks such as 3D shape classification and segmentation. Experiments are conducted on standard datasets, including synthetic and real scanned ones, and our PSNet achieves state-of-the-art accuracies for shape classification, as well as competitive results for shape segmentation compared with previous methods.

The impact of neoadjuvant chemotherapy on low anterior resection syndrome after rectal cancer resection: A 6 Months longitudinal follow-up.

BACKGROUND/OBJECTIVE: Neoadjuvant radiotherapy plays a vital role in rectal cancer treatment, but impairs postoperative bowel function, leading to low anterior resection syndrome (LARS). Neoadjuvant chemotherapy alone might avoid the negative effect of radiotherapy on bowel function. This study aims to assess the impact of neoadjuvant chemotherapy on LARS and the development of LARS over the first 6 months after surgery. METHODS: Rectal cancer patients were prospectively recruited during June 30, 2018 and December 24, 2019. Bowel function was assessed by the LARS score, which was taken at 1 month, 3 months, and 6 months after surgery via phone call interview. Patients were divided into two groups based on whether they received neoadjuvant chemotherapy (group A) or not (group B). RESULTS: A total of 97 patients were included in the analysis. There was no significant difference between the LARS scores at 1 month, 3 months, and 6 months of both groups. The LARS score at 6 months showed a significant decrease from that of 1 month and 3 months in group B (P < 0.05, P < 0.01) and in all patients (P < 0.05, P = 0.001), and significant difference was found between the LARS scores in group A at the three timepoints (P < 0.05). No significant difference was found between the scores at 1 month and 3 months in both groups and in all patients. CONCLUSION: Neoadjuvant chemotherapy alone did not have a negative impact on LARS. The bowel function after surgery started to show significant improvement at 6 months after surgery.

DRCNN: Dynamic Routing Convolutional Neural Network for Multi-View 3D Object Recognition.

3D object recognition is one of the most important tasks in 3D data processing, and has been extensively studied recently. Researchers have proposed various 3D recognition methods based on deep learning, among which a class of view-based approaches is a typical one. However, in the view-based methods, the commonly used view pooling layer to fuse multi-view features causes a loss of visual information. To alleviate this problem, in this paper, we construct a novel layer called Dynamic Routing Layer (DRL) by modifying the dynamic routing algorithm of capsule network, to more effectively fuse the features of each view. Concretely, in DRL, we use rearrangement and affine transformation to convert features, then leverage the modified dynamic routing algorithm to adaptively choose the converted features, instead of ignoring all but the most active feature in view pooling layer. We also illustrate that the view pooling layer is a special case of our DRL. In addition, based on DRL, we further present a Dynamic Routing Convolutional Neural Network (DRCNN) for multi-view 3D object recognition. Our experiments on three 3D benchmark datasets show that our proposed DRCNN outperforms many state-of-the-arts, which demonstrates the efficacy of our method.

Second-order Spectral Transform Block for 3D Shape Classification and Retrieval.

In this paper, we propose a novel network block, dubbed as second-order spectral transform block, for 3D shape retrieval and classification. This network block generalizes the second-order pooling to 3D surface by designing a learnable non-linear transform on the spectrum of the pooled descriptor. The proposed block consists of following two components. First, the second-order average (SO-Avr) and max-pooling (SOMax) operations are designed on 3D surface to aggregate local descriptors, which are shown to be more discriminative than the popular average-pooling or max-pooling. Second, a learnable spectral transform parameterized by mixture of power function is proposed to perform non-linear feature mapping in the space of pooled descriptors, i.e., manifold of symmetric positive definite matrix for SO-Avr, and space of symmetric matrix for SOMax. The proposed block can be plugged into existing network architectures to aggregate local shape descriptors for boosting their performance. We apply it to a shallow network for nonrigid 3D shape analysis and to existing networks for rigid shape analysis, where it improves the first-tier retrieval accuracy by 7.2% on SHREC'14 Real dataset and achieves state-of-the-art classification accuracy on ModelNet40. As an extension, we apply our block to 2D image classification, showing its superiority compared with traditional second-order pooling methods. We also provide theoretical and experimental analysis on stability of the proposed second-order spectral transform block.