Search

Non-maximum suppression (NMS) is an essential post-processing step for object detection. The de-facto standard for NMS, namely GreedyNMS, is not parallelizable and could thus be the performance bottleneck in object detection pipelines. MaxpoolNMS is introduced as a fast and parallelizable alternative to GreedyNMS. However, MaxpoolNMS is only capable of replacing the GreedyNMS at the first stage of two-stage detectors like Faster R-CNN. To address this issue, we observe that MaxpoolNMS employs the process of box coordinate discretization followed by local score argmax calculation, to discard the nested-loop pipeline in GreedyNMS to enable parallelizable implementations. In this paper, we introduce a simple Relationship Recovery module and a Pyramid Shifted MaxpoolNMS module to improve the above two stages, respectively. With these two modules, our PSRR-MaxpoolNMS is a generic and parallelizable approach, which can completely replace GreedyNMS at all stages in all detectors. Furthermore, we extend PSRR-MaxpoolNMS to the more powerful PSRR-MaxpoolNMS++. As for box coordinate discretization, we propose Density-based Discretization for better adherence to the target density of the suppression. As for local score argmax calculation, we propose an Adjacent Scale Pooling scheme for mining out the duplicated box pairs more accurately and efficiently. Extensive experiments demonstrate that both our PSRR-MaxpoolNMS and PSRR-MaxpoolNMS++ outperform MaxpoolNMS by a large margin. Additionally, PSRR-MaxpoolNMS++ not only surpasses PSRR-MaxpoolNMS but also attains competitive accuracy and much better efficiency when compared with GreedyNMS. Therefore, PSRR-MaxpoolNMS++ is a parallelizable NMS solution that can effectively replace GreedyNMS at all stages in all detectors.

Deep neural networks (DNNs) have been widely used in many artificial intelligence (AI) tasks. However, deploying them brings significant challenges due to the huge cost of memory, energy, and computation. To address these challenges, researchers have developed various model compression techniques such as model quantization and model pruning. Recently, there has been a surge in research on compression methods to achieve model efficiency while retaining performance. Furthermore, more and more works focus on customizing the DNN hardware accelerators to better leverage the model compression techniques. In addition to efficiency, preserving security and privacy is critical for deploying DNNs. However, the vast and diverse body of related works can be overwhelming. This inspires us to conduct a comprehensive survey on recent research toward the goal of high-performance, cost-efficient, and safe deployment of DNNs. Our survey first covers the mainstream model compression techniques, such as model quantization, model pruning, knowledge distillation, and optimizations of nonlinear operations. We then introduce recent advances in designing hardware accelerators that can adapt to efficient model compression approaches. In addition, we discuss how homomorphic encryption can be integrated to secure DNN deployment. Finally, we discuss several issues, such as hardware evaluation, generalization, and integration of various compression approaches. Overall, we aim to provide a big picture of efficient DNNs from algorithm to hardware accelerators and security perspectives.

BmK DKK13 (DKK13) is a mutated recombinant peptide, which has a significant antinociception in a rat model of the inflammatory pain. The purpose of this study was to evaluate the antinociceptive effect of DKK13 on trigeminal neuralgia (TN) in rats. Male Sprague-Dawley (SD) rats were treated with the chronic constriction injury of the infraorbital nerve (IoN-CCI) model to induce stable symptoms of TN. DKK13 (1.0 mg/kg, 2.0 mg/kg and 4.0 mg/kg, i.v.) or morphine (4.0 mg/kg, i.v.) was administered by tail vein once on day 14 after IoN-CCI injury. Behavioral tests, electrophysiology and western blotting were performed to investigate the role and underlying mechanisms of DKK13 on IoN-CCI model. Behavioral test results showed that DKK13 could significantly increase the mechanical pain and thermal radiation pain thresholds of IoN-CCI rats and inhibit the asymmetric spontaneous pain scratching behavior. Electrophysiological results showed that DKK13 could significantly reduce the current density of Nav1.8 in the ipsilateral side of trigeminal ganglion (TG) neurons in IoN-CCI rats, and the steady-state activation and inactivation curves of Nav1.8 shifted, respectively, to the direction of hyperpolarization and depolarization. Western blotting results showed that DKK13 significantly reduced the expression of Nav1.8 and the phosphorylation levels of key proteins of MAPKs/CREB pathway in TG tissues of IoN-CCI rats. In brief, DKK13 has a significant antinociceptive effect on IoN-CCI rats, which may be achieved by changing the dynamic characteristics of Nav1.8 channel and regulating the protein phosphorylation in MAPKs/CREB pathway., (© 2022. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.)

Atorvastatin ester (Ate) is a structural trim of atorvastatin that can regulate hyperlipidemia. The purpose of this study was to evaluate the lipid-lowering effect of Ate. Male Sprague Dawley (SD) rats were fed a high-fat diet for seven months and used as a hyperlipidemia model. The lipid level and liver function of the hyperlipidemia rats were studied by the levels of TG, TC, LDL, HDL, ALT, and AST in serum after intragastric administration with different doses of Ate. HE staining was used to observe the pathological changes of the rat liver and gastrocnemius muscle. The lipid deposits in the liver of rats were observed by staining with ORO. The genes in the rat liver were sequenced by RNA-sequencing. The results of the RNA-sequencing were further examined by qRT-PCR and western blotting. Biochemical test results indicated that Ate could obviously improve the metabolic disorder and reduce both the ALT and AST levels in serum of the hyperlipidemia rats. Pathological results showed that Ate could improve HFD-induced lipid deposition and had no muscle toxicity. The RNA-sequencing results suggested that Ate affected liver lipid metabolism and cholesterol, metabolism in the hyperlipidemia-model rats may vary via the PPAR-signaling pathway. The western blotting and qRT-PCR results demonstrated the Ate-regulated lipid metabolism in the hyperlipidemia model through the PPAR-signaling pathway and HMGCR expression. In brief, Ate can significantly regulate the blood lipid level of the model rats, which may be achieved by regulating the PPAR-signaling pathway and HMGCR gene expression.

Methamphetamine (METH) is a highly neurotoxic psychoactive substance that can directly damage the central nervous system through prolonged use. Oxytocin (OT) has attracted much attention because of its neuroprotective effect. The purpose of this study was to investigate whether OT is neuroprotective against METH-induced damage in rat hippocampal neurons. Our results revealed that pre-incubation with OT significantly prevented the damage of METH to hippocampal neurons, including the decrease of mitochondrial membrane potential and the increase of ROS (reactive oxygen species). OT pre-incubation attenuated the up-regulation of Cleaved-Caspase-3 expression and the down-regulation of Bcl-2/Bax expression induced by METH. Pre-incubation with OT prevented the decrease in oxytocin receptor density and P-CREB (phosphorylation of cAMP-response element binding) expression induced by METH in rat hippocampal neurons. Moreover, Pre-incubation of atosiban (ATO) significantly prevented these changes. In conclusion, our study proved that pre-administration of OT could significantly attenuate hippocampal neuron apoptosis induced by METH. Oxytocin receptor activation is involved in the preventive effect of OT on METH-induced apoptosis in rat hippocampal neurons., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2021 Li, Wang, Wang, Chen, Bai, Ban and Wu.)

Purpose: We conducted preclinical experiments and phase I clinical trial to investigate the safety, pharmacokinetics (PK) and antitumour effects of GT0918 in castration-resistant prostate cancer (CRPC)., Experimental Design: An androgen receptor (AR) competitive binding assay was performed, followed by evaluation of GT0918 on AR protein expression. The efficacy of GT0918 was investigated in a castration-resistant xenograft model. A phase I dose-escalation study of GT0918 in CRPC was also carried out to evaluate its safety, PK and antitumour efficacy., Results: GT0918 was demonstrated to inhibit the binding of androgen to AR more potently than MDV3100, and to effectively reduce the AR protein level. GT0918 inhibited the transcriptional activity of wild-type AR and AR with clinically relevant ligand-binding domain mutations. Furthermore, GT0918 significantly inhibited the growth of prostate cancer. A total of 16 patients was treated with GT0918 at five dose levels. Among these 16 patients, 10 and 2 patients, respectively, completed a three-cycle and six-cycle treatment, in which MTD was not reached. All the treatment-related adverse events were grade I, including hypercholesterolemia, hypertriglyceridemia, fatigue and anaemia. PK parameters showed that drug exposure increased with dose proportionally from 50 to 300 mg and a saturation was observed between 300 and 400 mg. PSA declines of ≥30% and ≥50% were, respectively, observed in six and two cases. All the 12 patients with metastatic soft tissue lesions confirmed stable disease., Conclusions: GT0918, a full AR antagonist without agonist effect, has high binding affinity to AR with AR protein down-regulation activity. GT0918 is demonstrated to be well tolerated with a favourable PK profile and exhibits promising antitumour activity in CRPC. CLINICALTRIALS: gov identifier CTR20150501., Competing Interests: Conflict of interest statement Ye Sun, Honghua Yan, Qianxiang Zhou, Qingqing Zhou, Meiyu Wang, Chunyun Chen are employees of Suzhou Kintor Pharmaceuticals. Ye Sun, Honghua Yan, Qianxiang Zhou, Qingqing Zhou, Meiyu Wang, Chunyun Chen also received personal fees from Suzhou Kintor Pharmaceuticals outside the submitted work. Tie Zhou, Weidong Xu, Wei Zhang, Xu Gao, Fubo Wang, Jianguo Hou, Shancheng Ren, Qing Yang, Bo Yang, Chuanliang Xu, Yinghao Sun are employees of Changhai Hospital, Second Military Medical University. They did not receive any personal fees, consulting fees and research funding from Suzhou Kintor Pharmaceuticals., (Copyright © 2020 Elsevier Ltd. All rights reserved.)

Human indoleamine 2,3-dioxygenase 1 (IDO1) is a heme-dependent enzyme with important roles in many cellular processes and is a potential target for drug discovery against cancer and other diseases. Crystal structures of IDO1 in complex with various inhibitors have been reported. Many of these crystals belong to the same crystal form and most of the reported structures have resolutions in the range 3.2-2.3 Å. Here, three new crystal forms of human IDO1 obtained by introducing a surface mutation, K116A/K117A, distant from the active site are reported. One of these crystal forms diffracted to 1.5 Å resolution and can be readily used for soaking experiments to determine high-resolution structures of IDO1 in complex with the substrate tryptophan or inhibitors that coordinate the heme. In addition, this mutant was used to produce crystals of a complex with an inhibitor that targets the apo form of the enzyme under the same conditions; the structure of this complex was determined at 1.7 Å resolution. Overall, this mutant represents a robust platform for determining the structures of inhibitor and substrate complexes of IDO1 at high resolution.