ZS-VAT: Learning Unbiased Attribute Knowledge for Zero-Shot Recognition Through Visual Attribute Transformer.

In zero-shot learning (ZSL), attribute knowledge plays a vital role in transferring knowledge from seen classes to unseen classes. However, most existing ZSL methods learn biased attribute knowledge, which usually results in biased attribute prediction and a decline in zero-shot recognition performance. To solve this problem and learn unbiased attribute knowledge, we propose a visual attribute Transformer for zero-shot recognition (ZS-VAT), which is an effective and interpretable Transformer designed specifically for ZSL. In ZS-VAT, we design an attribute-head self-attention (AHSA) that is capable of learning unbiased attribute knowledge. Specifically, each attribute head in AHSA first transforms the local features into attribute-reinforced features and then accumulates the attribute knowledge from all corresponding reinforced features, reducing the mutual influence between attributes and avoiding information loss. AHSA finally preserves unbiased attribute knowledge through attribute embeddings. We also propose an attribute fusion model (AFM) that learns to recover the correct category knowledge from the attribute knowledge. In particular, AFM takes all features from AHSA as input and generates global embeddings. We carried out experiments to demonstrate that the attribute knowledge from AHSA and the category knowledge from AFM are able to assist each other. During the final semantic prediction, we combine the attribute embedding prediction (AEP) and global embedding prediction (GEP). We evaluated the proposed scheme on three benchmark datasets. ZS-VAT outperformed the state-of-the-art generalized ZSL (GZSL) methods on two datasets and achieved competitive results on the other dataset.

Point Cloud Registration-Driven Robust Feature Matching for 3-D Siamese Object Tracking.

Learning robust feature matching between the template and search area is crucial for 3-D Siamese tracking. The core of Siamese feature matching is how to assign high feature similarity to the corresponding points between the template and the search area for precise object localization. In this article, we propose a novel point cloud registration-driven Siamese tracking framework, with the intuition that spatially aligned corresponding points (via 3-D registration) tend to achieve consistent feature representations. Specifically, our method consists of two modules, including a tracking-specific nonlocal registration (TSNR) module and a registration-aided Sinkhorn template-feature aggregation module. The registration module targets the precise spatial alignment between the template and the search area. The tracking-specific spatial distance constraint is proposed to refine the cross-attention weights in the nonlocal module for discriminative feature learning. Then, we use the weighted singular value decomposition (SVD) to compute the rigid transformation between the template and the search area and align them to achieve the desired spatially aligned corresponding points. For the feature aggregation model, we formulate the feature matching between the transformed template and the search area as an optimal transport problem and utilize the Sinkhorn optimization to search for the outlier-robust matching solution. Also, a registration-aided spatial distance map is built to improve the matching robustness in indistinguishable regions (e.g., smooth surfaces). Finally, guided by the obtained feature matching map, we aggregate the target information from the template into the search area to construct the target-specific feature, which is then fed into a CenterPoint-like detection head for object localization. Extensive experiments on KITTI, NuScenes, and Waymo datasets verify the effectiveness of our proposed method.

Efficient 3D Point Cloud Feature Learning for Large-Scale Place Recognition.

Point cloud based retrieval for place recognition is still a challenging problem since the drastic appearance changes of scenes due to seasonal or artificial changes in the environments. Existing deep learning based global descriptors for the retrieval task usually consume a large amount of computational resources ( e.g ., memory), which may not be suitable for the cases of limited hardware resources. In this paper, we develop an efficient point cloud learning network (EPC-Net) to generate global descriptors of point clouds for place recognition. While obtaining good performance, it can greatly reduce computational memory and inference time. First, we propose a lightweight but effective neural network module, called ProxyConv, to aggregate the local geometric features of point clouds. We leverage the adjacency matrix and proxy points to simplify the original edge convolution for lower memory consumption. Then, we design a lightweight grouped VLAD network to form global descriptors for retrieval. Compared with the original VLAD network, we propose a grouped fully connected layer to decompose the high-dimensional vectors into a group of low-dimensional vectors, which can reduce the number of parameters of the network and maintain the discrimination of the feature vector. Finally, we further develop a simple version of EPC-Net, called EPC-Net-L, which consists of two ProxyConv modules and one max pooling layer to aggregate global descriptors. By distilling the knowledge from EPC-Net, EPC-Net-L can obtain discriminative global descriptors for retrieval. Extensive experiments on the Oxford dataset and three in-house datasets demonstrate that our method achieves good results with lower parameters, FLOPs, GPU memory, and shorter inference time. Our code is available at https://github.com/fpthink/EPC-Net.

[The progress of novel drug delivery systems].

The development of pharmaceuticals has been providing many kinds of novel drug delivery systems, which are important for improving therapeutic effect and one of the most important fields in pharmaceutics. According to their application, we can generally divide the novel drug delivery systems into three categories: quickly performed drug delivery system, long-term drug delivery system and high effective drug delivery system. Some diseases, such as asthma, angina pectoris and migraine, require therapeutics urgently, and the drugs have to be absorbed in several minutes. Therefore, quickly performed drug delivery systems are developed, such as oral disintegrating tablets and nasal spray. For normal tablets and capsules, especially the drugs with short blood half life, the drug concentration in blood shows obvious peak-valley phenomenon, which reduces the therapeutic effect and requires multiple administration. To solve this problem, sustained drug release system was developed, which could release the drugs slowly and sustainably even in zero-order kinetics. The pulse drug delivery system was developed that can delayed and pulsed release drug for one or several times. This system is especially useful in the management of asthma and heart disease, which are often found in midnight or early morning when patients are in bed. Transdermal drug delivery system could release drugs sustainably and deliver the drugs through skin to blood circulation, providing long term activity. The water-insoluble drugs are difficult for pharmaceutical development, thus many methods were developed to improve the solubility and bioavailability of drugs. Although biopharmaceuticals are important for disease treatment, the application shadows by the poor stability and low bioavailability. Thus the biopharmaceutical delivery system was developed, which mainly focused on structure modification and encapsulation by carriers. Considering therapeutic effect requires interaction between drugs and their targets, it is important to deliver drugs to their targets. Therefore, targeting delivery systems were developed, which mainly based on the nanoparticles. Furthermore, on-demand release drug delivery systems are also developed with the property of environment-triggered drug release. In conclusion, the novel drug delivery systems were reviewed in this study.

[The development of novel tumor targeting delivery strategy].

Tumor is one of the most serious threats for human being. Although many anti-tumor drugs are approved for clinical use, the treatment outcome is still modest because of the poor tumor targeting efficiency and low accumulation in tumor. Therefore, it is important to deliver anti-tumor drug into tumor efficiently, elevate drug concentration in tumor tissues and reduce the drug distribution in normal tissues. And it has been one of the most attractive directions of pharmaceutical academy and industry. Many kinds of strategies, especially various nanoparticulated drug delivery systems, have been developed to address the critical points of complex tumor microenvironment, which are partially or mostly satisfied for tumor treatment. In this paper, we carefully reviewed the novel targeting delivery strategies developed in recent years. The most powerful method is passive targeting delivery based on the enhanced permeability and retention(EPR) effect, and most commercial nanomedicines are based on the EPR effect. However, the high permeability and retention require different particle sizes, thus several kinds of size-changeable nanoparticles are developed, such as size reducible particles and assemble particles, to satisfy the controversial requirement for particle size and enhance both tumor retention and penetration. Surface charge reversible nanoparticles also shows a high efficiency because the anionic charge in blood circulation and normal organs decrease the unintended internalization. The charge can change into positive in tumor microenvironment, facilitating drug uptake by tumor cells. Additionally, tumor microenvironment responsive drug release is important to decrease drug side effect, and many strategies are developed, such as p H sensitive release and enzyme sensitive release. Except the responsive nanoparticles, shaping tumor microenvironment could attenuate the barriers in drug delivery, for example, decreasing tumor collagen intensity and normalizing tumor microvessels to decrease the internal fluid pressure. All these strategies could enhance the accumulation and penetration of nanoparticles into tumor, leading to a homogenous distribution of drugs in tumor. To enhance the internalization by specific cells, active targeting delivery strategies are developed. There were many surface markers, receptors or carriers overexpressed on specific kinds of cells, thus the corresponding ligands were utilized to mediate active targeting to certain cells, including tumor cells, cancer stem cells, tumor neovasculatures, tumor associated macrophages and other tumor stroma cells. Targeting more than one cell type may provide an improved antitumor effect. Although these passive and active targeting strategies all have promising outcome in the treatment of tumor, some shortages are still unaddressed, such as the specificity of responsive is not good enough, and the active targeting may be diminished by the protein corona. Thus more research is required to promote the drug delivery study.

[The construction of cell-penetrating peptide R8 and pH sensitive cleavable polyethylene glycols co-modified liposomes].

The purpose of the study is to construct R8 peptide (RRRRRRRR) and pH sensitive polyethylene glycols (PEG) co-modified liposomes (Cl-Lip) and utilize them in breast cancer treatment. The co-modified liposomes were prepared with soybean phospholipid, cholesterol, DSPE-PEG2K-R8 and PEG5K-Hz-PE (pH sensitive PEG). The size and zeta potential of Cl-Lip were also characterized. The in vitro experiment demonstrated that the Cl-Lip had high serum stability in 50% fetal bovine serum. The cellular uptake of Cl-Lip under different pre-incubated conditions was evaluated on 4T1 cells. And the endocytosis pathway, lysosome escape ability and tumor spheroid penetration ability were also evaluated. The results showed the particle size of the Cl-Lip was (110.4 ± 5.2) nm, PDI of the Cl-Lip was 0.207 ± 0.039 and zeta potential of the Cl-Lip was (-3.46 ± 0.05) mV. The cellular uptake of Cl-Lip on 4T1 cells was pH sensitive, as the cellular uptake of Cl-Lip pre-incubated in pH 6.0 was higher than that of pH 7.4 under each time point. The main endocytosis pathways of Cl-Lip under pH 6.0 were micropinocytosis and energy-dependent pathway. At the same time, the Cl-Lip with pre-incubation in pH 6.0 had high lysosome escape ability and high tumor spheroid penetration ability. All the above results demonstrated that the Cl-Lip we constructed had high pH sensitivity and is a promising drug delivery system.

Polyethylene glycol modification decreases the cardiac toxicity of carbonaceous dots in mouse and zebrafish models.

AIM: Carbonaceous dots (CDs), which have been used for diagnosis, drug delivery and gene delivery, are accumulated in heart at high concentrations. To improve their biocompatibility, polyethylene glycol-modified CDs (PEG-CDs) were prepared. In this study we compared the cardiac toxicity of CDs and PEG-CDs in mouse and zebrafish models. METHODS: Mice were intravenously treated with CDs (size: 4.9 nm, 5 mg·kg(-1)·d(-1)) or PEG-CDs (size: 8.3 nm, 5 mg·kg(-1)·d(-1)) for 21 d. Their blood biochemistry indices, ECG, and histological examination were examined for evaluation of cardiac toxicity. CDs or PEG-CDs was added in incubator of cmlc2 transgenic Zebrafish embryos at 6 hpf, and the shape and size of embryos' hearts were observed at 48 hpf using a fluorescent microscope. Furthermore, whole-mount in situ hybridization was used to examine the expression of early cardiac marker gene (clml2) at 48 hpf. RESULTS: Administration of CDs or PEG-CDs in mice caused mild, but statistically insignificant reduction in serum creatine kinase (CK) and lactate dehydrogenase (LDH) levels detected at 7 d, which were returned to the respective control levels at 21 d. Neither CDs nor PEG-CDs caused significant changes in the morphology of heart cells. Administration of CDs, but not PEG-CDs, in mice caused marked increase of heart rate. Both CDs and PEG-CDs did not affect other ECG parameters. In the zebrafish embryos, addition of CDs (20 µg/mL) caused heart development delay, whereas addition of CDs (80 µg/mL) led to heart malformation. In contrast, PEG-CDs caused considerably small changes in heart development, which was consistent with the results from the in situ hybridization experiments. CONCLUSION: CDs causes greater cardiac toxicity, especially regarding heart development. Polyethylene glycol modification can attenuate the cardiac toxicity of CDs.

[Cell penetrating peptide TAT and brain tumor targeting peptide T7 dual modified liposome preparation and in vitro targeting evaluation].

The purpose of this study is to prepare T7 and TAT dual modified liposomes (T7-TAT-LIP) to penetrate through blood brain barrier and target to brain tumor cells. The liposomes were prepared with CFPE, T7 modified PEG-DSPE, TAT modified PEG-DSPE, soybean phospholipid, PEG-DSPE and cholesterol. The CFPE was used to track the cellular uptake efficiency. The density of T7 and TAT and the length of PEG were optimized, and then the liposomes were characterized by particle size, zeta potential, morphology and stability. Afterwards, the cellular uptake by bEnd.3 and C6 cells were evaluated. The results showed that the optimized parameters were 6% of T7, 0.5% of TAT, the molecular weight of PEG for T7 was 2000 and the molecular weight of PEG for TAT was 1000. After optimization, the particle size of T7-TAT-LIP was 118 nm, the zeta potential was -6.32 mV and the particles were spherical. The turbidity and particle size of liposomes were not obviously changed after 24 h incubation in PBS at 37 °C. The particle size and polydispersity index were also stable during 1 month incubation at 4-8 °C. The cellular uptake by both bEnd.3 and C6 cells of T7-TAT-LIP was higher than that of T7 or TAT modified liposomes, suggesting dual modified liposomes possessed better blood brain barrier targeting ability and brain tumor targeting ability than the single ligand modified liposomes.

[Fluorescent carbon dots and the application in biomedicine].

As a new type of carbon nanomaterials, fluorescent carbon dots (fluorescent CDs) have many advantages when compared with the traditional fluorescent probes. They are photoluminescence stable and resistance to photo bleaching. Moreover, they are excellent in biocompatibility, low-toxic and easy to modify. All these above make them a promising optical image material as a probe in optical image. This article reviews structure, the common carbon sources, the preparation methods, and the light-emitting principles of the carbon dots. We also introduce the research progress of fluorescent carbon dots in biomedicine, and the problems need to be resolved in the study of fluorescent CDs.

Lapatinib-incorporated lipoprotein-like nanoparticles: preparation and a proposed breast cancer-targeting mechanism.

AIM: Lapatinib is a dual inhibitor of EGFR and human epidermal growth factor receptor 2 (HER2), and used to treat advanced breast cancer. To overcome its poor water solubility, we constructed lapatinib-incorporated lipoprotein-like nanoparticles (LTNPs), and evaluated the particle characteristics and possible anti-breast cancer mechanisms. METHODS: LTNPs (lapatinib bound to albumin as a core, and egg yolk lecithin forming a lipid corona) were prepared. The particle characteristics were investigated using transmission electron microscopy (TEM) and atomic force microscopy (AFM). The uptake and subcellular localization of LTNPs, as well as the effects of LTNPs on cell cycle were examined in BT-474 human breast cancer cells in vitro. Mice bearing BT-474 subcutaneous xenograft were intravenously injected with coumarin-6 loaded LTNPs (30 mg/kg) to study the targeting mechanisms in vivo. RESULTS: The LTNPs particles were generally spherical but flexible under TEM and AFM, and approximately 62.1 nm in size with a zeta potential of 22.80 mV. In BT-474 cells, uptake of LTNPs was mediated by endosomes through energy-dependent endocytosis involving clathrin-dependent pinocytosis and macropinocytosis, and they could effectively escape from endosomes to the cytoplasm. Treatment of BT-474 cells with LTNPs (20 µg/mL) induced a significant cell arrest at G0/G1 phase compared with the same concentration of lapatinib suspension. In mice bearing BT-474 xenograft, intravenously injected LTNPs was found to target and accumulate in tumors, and colocalized with HER2 and SPRAC (secreted protein, acidic and rich in cysteine). CONCLUSION: LTNPs can be taken up into breast cancer cells through specific pathways in vitro, and targeted to breast cancer xenograft in vivo via enhanced permeability and retention effect and SPARC.

Facile fabrication and upconversion luminescence enhancement of LaF3:Yb3+/Ln3+@SiO2 (Ln = Er, Tm) nanostructures decorated with Ag nanoparticles.

A novel hybrid nanostructure, that is a Ag nanoparticle decorated LaF(3):Yb(3+)/Ln(3+)@SiO(2) nanosphere (Ln=Er, Tm), was constructed by a facile strategy, and characterized by XRD, TEM, FTIR, XPS and UV-vis-NIR absorption. Obvious spectral broadening and red-shift on the surface plasmon resonance were obtained by adjusting the size and configuration of Ag nanoparticles. Effective upconversion luminescence enhancements for Er(3+) and Tm(3+) containing samples were obtained. It is suggested that the luminescence enhancement results from both the excitation and emission processes, and the configuration of the studied hybrid nanostructure is an efficient system to enhance the luminescence emission of rare earth doped nanomaterials. It is believed that the enhancement from the hybrid nanostructure will find great potential in the development of photovoltaic solar cells.

Effect of lactoferrin- and transferrin-conjugated polymersomes in brain targeting: in vitro and in vivo evaluations.

AIM: To evaluate the effect of lactoferrin (Lf) and transferrin (Tf) in brain targeting. METHODS: Polymersomes (PSs), employed as vectors, were conjugated with Lf or Tf and were characterized by morphology, particle size, zeta potential, and surface densities of the Lf or Tf molecules. In vitro uptake of Lf-PS and Tf-PS by bEnd.3 cells was investigated using coumarin-6 as a fluorescent probe. In vivo tissue distribution and pharmacokinetics of (125)I-Lf-PS and (125)I-Tf-PS were also examined. RESULTS: The mean particle size of PS, Lf-PS, and Tf-PS was around 150 nm and the zeta potential of the PSs was about -20 mV. Less than 0.12% of the coumarin was released from coumarin-6-loaded PS in 84 h indicating that coumarin-6 was an accurate probe for the PSs' behavior in vitro. It was shown that the uptake of Lf-PS and Tf-PS by bEnd.3 cells was time-, temperature-, and concentration-dependent. Both Lf and Tf could increase the cell uptake of PSs at 37 degrees C, but the uptake of Tf-PS was significantly greater than that of Lf-PS. In vivo tissue distribution and pharmacokinetics in mice revealed higher brain uptake and distribution of Tf-PS than Lf-PS, which was in accordance with in vitro uptake results. The drug targeting index (DTI) of Tf-PS with regard to Lf-PS was 1.51. CONCLUSION: Using a PS as the delivery vector and bEnd.3 cells as the model of the blood-brain barrier (BBB), Tf was more effective than Lf in brain targeting.