Robust Fine-Grained Visual Recognition With Neighbor-Attention Label Correction.

Existing deep learning methods for fine-grained visual recognition often rely on large-scale, well-annotated training data. Obtaining fine-grained annotations in the wild typically requires concentration and expertise, such as fine category annotation for species recognition, instance annotation for person re-identification (re-id) and dense annotation for segmentation, which inevitably leads to label noise. This paper aims to tackle label noise in deep model training for fine-grained visual recognition. We propose a Neighbor-Attention Label Correction (NALC) model to correct labels during the training stage. NALC samples a training batch and a validation batch from the training set. It hence leverages a meta-learning framework to correct labels in the training batch based on the validation batch. To enhance the optimization efficiency, we introduce a novel nested optimization algorithm for the meta-learning framework. The proposed training procedure consistently improves label accuracy in the training batch, consequently enhancing the learned image representation. Experimental results demonstrate that our method significantly increases label accuracy from 70% to over 98% and outperforms recent approaches by up to 13.4% in mean Average Precision (mAP) on various fine-grained image retrieval (FGIR) tasks, including instance retrieval on CUB200 and person re-id on Market1501. We also demonstrate the efficacy of NALC on noisy semantic segmentation datasets generated from Cityscapes, where it achieves a significant 7.8% improvement in mIOU score. NALC also exhibits robustness to different types of noise, including simulated noise such as Asymmetric, Pair-Flip, and Pattern noise, as well as practical noisy labels generated by tracklets and clustering.

Graph-based social relation inference with multi-level conditional attention.

Social relation inference intrinsically requires high-level semantic understanding. In order to accurately infer relations of persons in images, one needs not only to understand scenes and objects in images, but also to adaptively attend to important clues. Unlike prior works of classifying social relations using attention on detected objects, we propose a MUlti-level Conditional Attention (MUCA) mechanism for social relation inference, which attends to scenes, objects and human interactions based on each person pair. Then, we develop a transformer-style network to achieve the MUCA mechanism. The novel network named as Graph-based Relation Inference Transformer (i.e., GRIT) consists of two modules, i.e., a Conditional Query Module (CQM) and a Relation Attention Module (RAM). Specifically, we design a graph-based CQM to generate informative relation queries for all person pairs, which fuses local features and global context for each person pair. Moreover, we fully take advantage of transformer-style networks in RAM for multi-level attentions in classifying social relations. To our best knowledge, GRIT is the first for inferring social relations with multi-level conditional attention. GRIT is end-to-end trainable and significantly outperforms existing methods on two benchmark datasets, e.g., with performance improvement of 7.8% on PIPA and 9.6% on PISC.

Tdrd3-null mice show post-transcriptional and behavioral impairments associated with neurogenesis and synaptic plasticity.

The Topoisomerase 3B (Top3b) - Tudor domain containing 3 (Tdrd3) protein complex is the only dual-activity topoisomerase complex that can alter both DNA and RNA topology in animals. TOP3B mutations in humans are associated with schizophrenia, autism and cognitive disorders; and Top3b-null mice exhibit several phenotypes observed in animal models of psychiatric and cognitive disorders, including impaired cognitive and emotional behaviors, aberrant neurogenesis and synaptic plasticity, and transcriptional defects. Similarly, human TDRD3 genomic variants have been associated with schizophrenia, verbal short-term memory and educational attainment. However, the importance of Tdrd3 in normal brain function has not been examined in animal models. Here we generated a Tdrd3-null mouse strain and demonstrate that these mice display both shared and unique defects when compared to Top3b-null mice. Shared defects were observed in cognitive behaviors, synaptic plasticity, adult neurogenesis, newborn neuron morphology, and neuronal activity-dependent transcription; whereas defects unique to Tdrd3-deficient mice include hyperactivity, changes in anxiety-like behaviors, olfaction, increased new neuron complexity, and reduced myelination. Interestingly, multiple genes critical for neurodevelopment and cognitive function exhibit reduced levels in mature but not nascent transcripts. We infer that the entire Top3b-Tdrd3 complex is essential for normal brain function, and that defective post-transcriptional regulation could contribute to cognitive and psychiatric disorders.

A dual-selective thermal emitter with enhanced subambient radiative cooling performance.

Radiative cooling is a zero-energy technology that enables subambient cooling by emitting heat into outer space (~3 K) through the atmospheric transparent windows. However, existing designs typically focus only on the main atmospheric transparent window (8-13 µm) and ignore another window (16-25 µm), under-exploiting their cooling potential. Here, we show a dual-selective radiative cooling design based on a scalable thermal emitter, which exhibits selective emission in both atmospheric transparent windows and reflection in the remaining mid-infrared and solar wavebands. As a result, the dual-selective thermal emitter exhibits an ultrahigh subambient cooling capacity (~9 °C) under strong sunlight, surpassing existing typical thermal emitters (≥3 °C cooler) and commercial counterparts (as building materials). Furthermore, the dual-selective sample also exhibits high weather resistance and color compatibility, indicating a high practicality. This work provides a scalable and practical radiative cooling design for sustainable thermal management.

Lateral hypothalamic glutamatergic inputs to VTA glutamatergic neurons mediate prioritization of innate defensive behavior over feeding.

The lateral hypothalamus (LH) is involved in feeding behavior and defense responses by interacting with different brain structures, including the Ventral Tegmental Area (VTA). Emerging evidence indicates that LH-glutamatergic neurons infrequently synapse on VTA-dopamine neurons but preferentially establish multiple synapses on VTA-glutamatergic neurons. Here, we demonstrated that LH-glutamatergic inputs to VTA promoted active avoidance, long-term aversion, and escape attempts. By testing feeding in the presence of a predator, we observed that ongoing feeding was decreased, and that this predator-induced decrease in feeding was abolished by photoinhibition of the LH-glutamatergic inputs to VTA. By VTA specific neuronal ablation, we established that predator-induced decreases in feeding were mediated by VTA-glutamatergic neurons but not by dopamine or GABA neurons. Thus, we provided evidence for an unanticipated neuronal circuitry between LH-glutamatergic inputs to VTA-glutamatergic neurons that plays a role in prioritizing escape, and in the switch from feeding to escape in mice.

Farnesoid X receptor activation protects against renal fibrosis via modulation of ß-catenin signaling.

OBJECTIVE: Activation of farnesoid X receptor (FXR), a bile acid nuclear receptor, may be implicated in the pathophysiology of diabetic nephropathy. We explored a possible role for FXR activation in preventing renal fibrosis in high fat diet (HFD)-fed mice. METHODS: We investigated the effects of HFD on mouse kidney and renal tubular epithelial cells both in vivo and in vitro, and observed the changes of FXR and ß-catenin pathway. FXR agonist was also used to alleviate this HFD-induced effect, and the interaction between FXR and ß-catenin was further verified. RESULTS: Mice were fed by a 60% kcal fat diet for 20 weeks developed the typical traits of metabolic syndrome with subsequent renal lipid accumulation and renal injury. Treatment with the FXR agonist CDCA or GW4064 decreased body weight, renal lipid accumulation, as well as renal injury. Moreover, renal ß-catenin signaling was activated and improved with FXR-agonist treatment in HFD-fed mice. To examine whether FXR affected ß-catenin signaling, and was involved in tubulo-interstitial fibrosis, we explored the FXR expression and function in ox-LDL induced-renal tubular injury. In rat proximal tubular epithelial cells (NRK-52E) stimulated by ox-LDL, FXR protein was decreased compared to control group, and phosphorylated (Ser675) ß-catenin was activated by ox-LDL in a dose- and time-dependent manner. Ox-LDL enhanced α-SMA and fibronectin expressions and reduced E-cadherin levels, whereas FXR agonism or FXR overexpression inhibited fibronectin and α-SMA expressions and restored E-cadherin. Moreover, FXR agonist treatment also decreased phosphorylated (Ser675) ß-catenin, nuclear translocation and ß-catenin-mediated transcription induced by ox-LDL in NRK-52E cells. We showed that FXR could bind with ß-catenin via the AF1 domain, and disrupt the assembly of the core ß-catenin/TCF4 complex. CONCLUSION: These experimental data suggest that FXR activation, via modulating ß-catenin signaling, may contribute to attenuating the development of lipid-mediated tubulo-interstitial fibrosis.

Measurable residual disease conversion rate with consolidation chemotherapy in acute myeloid leukemia.

The rate of MRD clearance in AML with standard consolidation chemotherapy is not well defined. A multi-institution retrospective analysis was performed on 107 consecutively treated AML patients in morphologic complete remission with detectable MRD post-induction therapy who received standard chemotherapy consolidation. In response to standard intermediate/high-dose cytarabine consolidation therapy, 26 of 60 patients (43.3%) with MRD threshold of detection of at least 0.1% converted to MRD-negative status (undetectable with assay used), and 6 of 47 patients (12.8%) with MRD threshold of detection > 0.1% converted to MRD-negative status. Multivariable logistic regression for patients with MRD threshold of detection of at least 0.1% showed that, when controlling for age, ELN risk category, dose of cytarabine, and use of a combination agent, treatment with 1 cycle of consolidation cytarabine versus ≥2 cycles decreased the odds of conversion of AML to MRD-negative (OR = 0.24, 95% CI 0.07-0.85, p = 0.03).

Distinct sub-second dopamine signaling in dorsolateral striatum measured by a genetically-encoded fluorescent sensor.

The development of genetically encoded dopamine sensors such as dLight has provided a new approach to measuring slow and fast dopamine dynamics both in brain slices and in vivo, possibly enabling dopamine measurements in areas like the dorsolateral striatum (DLS) where previously such recordings with fast-scan cyclic voltammetry (FSCV) were difficult. To test this, we first evaluated dLight photometry in mouse brain slices with simultaneous FSCV and found that both techniques yielded comparable results, but notable differences in responses to dopamine transporter inhibitors, including cocaine. We then used in vivo fiber photometry with dLight in mice to examine responses to cocaine in DLS. We also compared dopamine responses during Pavlovian conditioning across the striatum. We show that dopamine increases were readily detectable in DLS and describe transient dopamine kinetics, as well as slowly developing signals during conditioning. Overall, our findings indicate that dLight photometry is well suited to measuring dopamine dynamics in DLS.

Lateral preoptic area glutamate neurons relay nociceptive information to the ventral tegmental area.

The ventral tegmental area (VTA) has been proposed to play a role in pain, but the brain structures modulating VTA activity in response to nociceptive stimuli remain unclear. Here, we demonstrate that the lateral preoptic area (LPO) glutamate neurons relay nociceptive information to the VTA. These LPO glutamatergic neurons synapsing on VTA neurons respond to nociceptive stimulation and conditioned stimuli predicting nociceptive stimulation and also mediate aversion. In contrast, LPO GABA neurons synapsing in the VTA mediate reward. By ultrastructural quantitative synaptic analysis, ex vivo electrophysiology, and functional neuroanatomy we identify a complex circuitry between LPO glutamatergic and GABAergic neurons and VTA dopaminergic, GABAergic, and glutamatergic neurons. We conclude that LPO glutamatergic neurons play a causal role in the processing of nociceptive stimuli and in relaying information about nociceptive stimuli. The pathway from LPO glutamatergic neurons to the VTA represents an unpredicted interface between peripheral nociceptive information and the limbic system.

Targeting the programmed cell death (PCD) signaling mechanism with natural substances for the treatment of diabetic cardiomyopathy (DCM).

Diabetic cardiomyopathy (DCM) is one of the severe complications of diabetes, characterized by structural and functional abnormalities in the hearts of diabetic patients without hypertension, coronary heart disease, or valvular heart disease. DCM can progress to heart failure, which is a significant cause of death in diabetic patients, but currently, there is no effective treatment available. Programmed cell death (PCD) is a genetically regulated form of cell death that includes apoptosis, autophagy, necroptosis, ferroptosis, and pyroptosis. PCD is essential for tissue homeostasis and normal development of the body. DCM is a complex condition, and abnormalities in the cascade of PCD signaling have been observed in its pathological process, suggesting that targeting PCD could be a potential therapeutic strategy. Studies have shown that natural substances can effectively modulate PCD to intervene in the treatment of DCM, and their use is safe. This review explores the role of different forms of PCD in the pathogenesis of DCM and summarizes the research progress in targeting PCD with natural substances to treat DCM. It can serve as a basis for further research and drug development to provide new treatment strategies for DCM patients.

AAformer: Auto-Aligned Transformer for Person Re-Identification.

In person re-identification (re-ID), extracting part-level features from person images has been verified to be crucial to offer fine-grained information. Most of the existing CNN-based methods only locate the human parts coarsely, or rely on pretrained human parsing models and fail in locating the identifiable nonhuman parts (e.g., knapsack). In this article, we introduce an alignment scheme in transformer architecture for the first time and propose the auto-aligned transformer (AAformer) to automatically locate both the human parts and nonhuman ones at patch level. We introduce the "Part tokens (PARTs)", which are learnable vectors, to extract part features in the transformer. A PART only interacts with a local subset of patches in self-attention and learns to be the part representation. To adaptively group the image patches into different subsets, we design the auto-alignment. Auto-alignment employs a fast variant of optimal transport (OT) algorithm to online cluster the patch embeddings into several groups with the PARTs as their prototypes. AAformer integrates the part alignment into the self-attention and the output PARTs can be directly used as part features for retrieval. Extensive experiments validate the effectiveness of PARTs and the superiority of AAformer over various state-of-the-art methods.

Evaluation of Open-Source Tools for Differential Privacy.

Differential privacy (DP) defines privacy protection by promising quantified indistinguishability between individuals who consent to share their privacy-sensitive information and those who do not. DP aims to deliver this promise by including well-crafted elements of random noise in the published data, and thus there is an inherent tradeoff between the degree of privacy protection and the ability to utilize the protected data. Currently, several open-source tools have been proposed for DP provision. To the best of our knowledge, there is no comprehensive study for comparing these open-source tools with respect to their ability to balance DP's inherent tradeoff as well as the use of system resources. This work proposes an open-source evaluation framework for privacy protection solutions and offers evaluation for OpenDP Smartnoise, Google DP, PyTorch Opacus, Tensorflow Privacy, and Diffprivlib. In addition to studying their ability to balance the above tradeoff, we consider discrete and continuous attributes by quantifying their performance under different data sizes. Our results reveal several patterns that developers should have in mind when selecting tools under different application needs and criteria. This evaluation survey can be the basis for an improved selection of open-source DP tools and quicker adaptation of DP.

Expression of the excitatory opsin ChRERα can be traced longitudinally in rat and nonhuman primate brains with PET imaging.

Optogenetics is a widely used technology with potential for translational research. A critical component of such applications is the ability to track the location of the transduced opsin in vivo. To address this problem, we engineered an excitatory opsin, ChRERα (hChR2(134R)-V5-ERα-LBD), that could be visualized using positron emission tomography (PET) imaging in a noninvasive, longitudinal, and quantitative manner. ChRERα consists of the prototypical excitatory opsin channelrhodopsin-2 (ChR2) and the ligand-binding domain (LBD) of the human estrogen receptor α (ERα). ChRERα showed conserved ChR2 functionality and high affinity for [18F]16α-fluoroestradiol (FES), an FDA-approved PET radiopharmaceutical. Experiments in rats demonstrated that adeno-associated virus (AAV)-mediated expression of ChRERα enables neural circuit manipulation in vivo and that ChRERα expression could be monitored using FES-PET imaging. In vivo experiments in nonhuman primates (NHPs) confirmed that ChRERα expression could be monitored at the site of AAV injection in the primary motor cortex and in long-range neuronal terminals for up to 80 weeks. The anatomical connectivity map of the primary motor cortex identified by FES-PET imaging of ChRERα expression overlapped with a functional connectivity map identified using resting state fMRI in a separate cohort of NHPs. Overall, our results demonstrate that ChRERα expression can be mapped longitudinally in the mammalian brain using FES-PET imaging and can be used for neural circuit modulation in vivo.

Efficient Vision Transformer via Token Merger.

Vision Transformers (ViTs) split an image into fixed-size patches as tokens. This strategy has succeeded in computer vision tasks, but introduces considerable tokens similar in semantics and appearances. This work proposes Token Merger to spot redundant tokens and merge them into a compact representation to accelerate ViTs. For each forward inference, the Token Merger first identifies meta tokens to represent meaningful cues of the image content, then adaptively merges similar tokens into a uniform one referring to meta tokens. To pursue a reasonable tradeoff between accuracy and efficiency, we further introduce learnable gates to adaptively decide the token merge ratios of different layers. As a generalizable module, Token Merger can be easily plugged into different layers of ViTs to boost their efficiency. Visualizations show that Token Merger progressively merges tokens and finally learns a compact set of tokens representing clear semantics. Compared with token pruning methods, Token Merger is more effective in preserving meaning contextual cues, thus performs and generalizes substantially better in different vision tasks. Extensive experiments and comparisons with other state-of-the-art downsampling methods also demonstrate its promising performance. For instance, it reduces 95% tokens and accelerates the inference speed by 62%. Meanwhile, the ImageNet classification accuracy only drops by 0.4%. The code will be available.

Impact of Thiotepa-Based Autologous Hematopoietic Cell Transplantation in Primary Central Nervous System Lymphoma in First Complete Remission: A University of California Hematologic Malignancies Consortium Retrospective Analysis.

BACKGROUND: The choice between nonmyeloablative chemotherapy (NMA-C) or autologous hematopoietic cell transplantation (autoHCT) as consolidation in primary central nervous system lymphoma (PCNSL), and timing of autoHCT differs among centers. We aimed to clarify these points. METHODS: We retrospectively analyzed PCNSL adult patients who received consolidation in CR1 or underwent autoHCT during their treatment course. Cohort A included those who underwent autoHCT in CR1, cohort B included those who underwent NMA-C in CR1, and cohort C included patients who underwent autoHCT in CR2+. We compared cohorts A and B, and cohorts A and C. The primary endpoint was overall survival (OS), and secondary endpoints were progression-free survival (PFS), treatment-related mortality (TRM) and cumulative incidence of relapse (CIR). RESULTS: 36 patients were included in cohort A, 30 in cohort B, and 14 in cohort C. The 5-year OS for cohorts A vs B and vs C were 90.7% vs 62.8% (P = .045) and vs 77.9% (P = .32), respectively. The 5-year PFS from diagnosis for cohorts A vs B was 87.8% vs 37.3% (P < .001). The 5-year PFS from autoHCT for cohorts A vs C was 87.6% vs 58.4% (P = .023). The 5-year TRM and CIR in cohorts A vs B was 9.4% vs 9.5% (P = .674), and 2.9% vs 53.2% (P < .001), respectively. The 5-year TRM and CIR in cohorts A vs C from the time of autoHCT was 9.5% vs 22.1% (P = .188), and 2.9% vs 19.5% (P = .104), respectively. CONCLUSION: Despite the limitations, thiotepa-based autoHCT in CR1 appears to improve outcomes in eligible patients with PCNSL.

Exploring the active ingredients and mechanism of Shenzhi Tongxin capsule against microvascular angina based on network pharmacology and molecular docking.

BACKGROUND: Microvascular angina (MVA) substantially threatens human health, and the Shenzhi Tongxin (SZTX) capsule demonstrates a remarkable cardioprotective effect, making it a potential treatment option for MVA. However, the precise mechanism of action for this medication remains unclear. This study utilized network pharmacology and molecular docking technology to investigate the active components and potential mechanisms underlying the efficacy of the SZTX capsule in alleviating MVA. METHODS: The main ingredients of the SZTX capsule, along with their targets proteins and potential disease targets associated with MVA, were extracted from public available databases. This study utilized the STRING database and Cytoscape 3.7.2 software to establish a protein-protein interaction network and determine key signaling pathway targets. Subsequently, the DAVID database was utilized to conduct Gene Ontology and the Kyoto Encyclopedia of Genes and Genomes analyses on the intersection targets. To further investigate the molecular interactions, Autodock and PyMOL software were employed to perform molecular docking and visualize the resulting outcomes. RESULTS: A total of 130 and 142 bioactive ingredients and intersection targets were identified respectively. Six core targets were obtained through protein-protein interaction network analysis. Gene Ontology enrichment analysis showed that 610 biological processes, 75 cellular components, and 92 molecular functions were involved. The results of Kyoto Encyclopedia of Genes and Genomes enrichment analyses indicated that SZTX capsule molecular mechanism in the treatment of MVA may be related to several pathways, including mitogen-activated protein kinases, PI3K-Akt, HIF-1, and others. The results of molecular docking showed that the 7 key active ingredients of SZTX capsule had good binding ability to 6 core proteins. CONCLUSION: SZTX capsule potentially exerts its effects by targeting multiple signaling pathways, including the mitogen-activated protein kinases signaling pathway, PI3K-Akt signaling pathway, and HIF-1 signaling pathway. This multi-target approach enables SZTX capsule to inhibit inflammation, alleviate oxidative stress, regulate angiogenesis, and enhance endothelial function.

Research progress of Traditional Chinese Medicine (TCM) in targeting inflammation and lipid metabolism disorder for arteriosclerosis intervention: A review.

Atherosclerosis (AS) is a chronic disease caused by inflammation and lipid deposition. Immune cells are extensively activated in the lesions, producing excessive pro-inflammatory cytokines, which accompany the entire pathological process of AS. In addition, the accumulation of lipid-mediated lipoproteins under the arterial intima is a crucial event in the development of AS, leading to vascular inflammation. Improving lipid metabolism disorders and inhibiting inflammatory reactions are the primary treatment methods currently used in medical practice to delay AS progression. With the development of traditional Chinese medicine (TCM), more mechanisms of action of the monomer of TCM, Chinese patent medicine, and compound prescription have been studied and explored. Research has shown that some Chinese medicines can participate in treating AS by targeting and improving lipid metabolism disorders and inhibiting inflammatory reactions. This review explores the research on Chinese herbal monomers, compound Chinese medicines, and formulae that improve lipid metabolism disorders and inhibit inflammatory reactions to provide new supplements for treating AS.

Flow Cytometry of Synaptoneurosomes (FCS) Reveals Increased Ribosomal S6 and Calcineurin Proteins in Activated Medial Prefrontal Cortex to Nucleus Accumbens Synapses.

Learning and behavior activate cue-specific patterns of sparsely distributed cells and synapses called ensembles that undergo memory-encoding engram alterations. While Fos is often used to label selectively activated cell bodies and identify neuronal ensembles, there is no comparable endogenous marker to label activated synapses and identify synaptic ensembles. For the purpose of identifying candidate synaptic activity markers, we optimized a flow cytometry of synaptoneurosome (FCS) procedure for assessing protein alterations in activated synapses from male and female rats. After injecting yellow fluorescent protein (YFP)-expressing adeno-associated virus into medial prefrontal cortex (mPFC) to label terminals in nucleus accumbens (NAc) of rats, we injected 20 mg/kg cocaine in a novel context (cocaine+novelty) to activate synapses, and prepared NAc synaptoneurosomes 0-60 min following injections. For FCS, we used commercially available antibodies to label presynaptic and postsynaptic markers synaptophysin and PSD-95 as well as candidate markers of synaptic activity [activity-regulated cytoskeleton protein (Arc), CaMKII and phospho-CaMKII, ribosomal protein S6 (S6) and phospho-S6, and calcineurin and phospho-calcineurin] in YFP-labeled synaptoneurosomes. Cocaine+novelty increased the percentage of S6-positive synaptoneurosomes at 5-60 min and calcineurin-positive synaptoneurosomes at 5-10 min. Electron microscopy verified that S6 and calcineurin levels in synaptoneurosomes were increased 10 min after cocaine+novelty. Pretreatment with the anesthetic chloral hydrate blocked cocaine+novelty-induced S6 and calcineurin increases in synaptoneurosomes, and novel context exposure alone (without cocaine) increased S6, both of which indicate that these increases were due to neural activity per se. Overall, FCS can be used to study protein alterations in activated synapses coming from specifically labeled mPFC projections to NAc.SIGNIFICANCE STATEMENT Memories are formed during learning and are stored in the brain by long-lasting molecular and cellular alterations called engrams formed within specific patterns of cue-activated neurons called neuronal ensembles. While Fos has been used to identify activated ensemble neurons and the engrams within them, we have not had a similar marker for activated synapses that can be used to identify synaptic engrams. Here we developed a procedure for high-throughput in-line analysis of flow cytometry of synaptoneurosome (FCS) and found that ribosomal S6 protein and calcineurin were increased in activated mPFC-NAc synapses. FCS can be used to study protein alterations in activated synapses within specifically labeled circuits.

Endoplasmic Reticulum Stress: A Key Regulator of Cardiovascular Disease.

The problems associated with economic development and social progress have led to an increase in the occurrence of cardiovascular diseases (CVDs), which affect the health of an increasing number of people and are a leading cause of disease and population mortality worldwide. Endoplasmic reticulum stress (ERS), a hot topic of interest for scholars in recent years, has been confirmed in numerous studies to be an important pathogenetic basis for many metabolic diseases and play an important role in maintaining physiological processes. The endoplasmic reticulum (ER) is a major organelle that is involved in protein folding and modification synthesis, and ERS occurs when several physiological and pathological factors allow excessive amounts of unfolded/misfolded proteins to accumulate. ERS often leads to initiation of the unfolded protein response (UPR) in a bid to re-establish tissue homeostasis; however, UPR has been documented to induce vascular remodeling and cardiomyocyte damage under various pathological conditions, leading to or accelerating the development of CVDs such as hypertension, atherosclerosis, and heart failure. In this review, we summarize the latest knowledge gained concerning ERS in terms of cardiovascular system pathophysiology, and discuss the feasibility of targeting ERS as a novel therapeutic target for the treatment of CVDs. Investigation of ERS has immense potential as a new direction for future research involving lifestyle intervention, the use of existing drugs, and the development of novel drugs that target and inhibit ERS.