Multi-Functional Regulation on Buried Interface for Achieving Efficient Triple-Cation Perovskite Solar Cells.

Mixed-cation perovskite solar cells (PSCs) have attracted much attention because of the advantages of suitable bandgap and stability. It is still a challenge to rationally design and modify the perovskite/tin oxide (SnO2) heterogeneous interface for achieving highly efficient and stable PSCs. Herein, a strategy of one-stone-for-three-birds is proposed to achieve multi-functional interface regulation via introducing N-Chlorosuccinimide (NCS) into the solution of SnO2: i) C═O functional group in NCS can induces strong binding affinity to uncoordinated defects (oxygen vacancies, free lead ions, etc) at the buried interface and passivate them; ii) incomplete in situ hydrolysis reactions can occur spontaneously and adjust the pH value of the SnO2 solution to achieve a more matchable energy level; iii) effectively releasing the residual stress of the underlying perovskite. As a result, a champion power conversion efficiency (PCE) of 24.74% is achieved with a device structure of ITO/SnO2/Perovskite/Spiro-OMeTAD/Ag, which is one of the highest values for cesium-formamidinium-methylammonium (CsFAMA) triple cation PSCs. Furthermore, the device without encapsulation can sustain 94.6% of its initial PCE after the storage at room temperature and relative humidity (RH) of 20% for 40 days. The research provides a versatile way to manipulate buried interface for achieving efficient and stable PSCs.

Learning Association Characteristics by Dynamic Hypergraph and Gated Convolution Enhanced Pairwise Attributes for Prediction of Disease-Related lncRNAs.

As the long non-coding RNAs (lncRNAs) play important roles during the incurrence and development of various human diseases, identifying disease-related lncRNAs can contribute to clarifying the pathogenesis of diseases. Most of the recent lncRNA-disease association prediction methods utilized the multi-source data about the lncRNAs and diseases. A single lncRNA may participate in multiple disease processes, and multiple lncRNAs usually are involved in the same disease process synergistically. However, the previous methods did not completely exploit the biological characteristics to construct the informative prediction models. We construct a prediction model based on adaptive hypergraph and gated convolution for lncRNA-disease association prediction (AGLDA), to embed and encode the biological characteristics about lncRNA-disease associations, the topological features from the entire heterogeneous graph perspective, and the gated enhanced pairwise features. First, the strategy for constructing hyperedges is designed to reflect the biological characteristic that multiple lncRNAs are involved in multiple disease processes. Furthermore, each hyperedge has its own biological perspective, and multiple hyperedges are beneficial for revealing the diverse relationships among multiple lncRNAs and diseases. Second, we encode the biological features of each lncRNA (disease) node using a strategy based on dynamic hypergraph convolutional networks. The strategy may adaptively learn the features of the hyperedges and formulate the dynamically evolved hypergraph topological structure. Third, a group convolutional network is established to integrate the entire heterogeneous topological structure and multiple types of node attributes within an lncRNA-disease-miRNA graph. Finally, a gated convolutional strategy is proposed to enhance the informative features of the lncRNA-disease node pairs. The comparison experiments indicate that AGLDA outperforms seven advanced prediction methods. The ablation studies confirm the effectiveness of major innovations, and the case studies validate AGLDA's ability in application for discovering potential disease-related lncRNA candidates.

Constructing Additives Synergy Strategy to Doctor-Blade Efficient CH3 NH3 PbI3 Perovskite Solar Cells under a Wide Range of Humidity from 45% to 82.

Perovskite solar cells (PSCs) have emerged as one of the most promising and competitive photovoltaic technologies, and doctor-blading is a facile and robust deposition technique to efficiently fabricate PSCs in large scale, especially matching with roll-to-roll process. Herein, it demonstrates the encouraging results of one-step, antisolvent-free doctor-bladed methylammonium lead iodide (CH3 NH3 PbI3, MAPbI3 ) PSCs under a wide range of humidity from 45% to 82%. A synergy strategy of ionic-liquid methylammonium acetate (MAAc) and molecular phenylurea additives is developed to modulate the morphology and crystallization process of MAPbI3 perovskite film, leading to high-quality MAPbI3 perovskite film with large-size crystal, low defect density, and ultrasmooth surface. Impressive power conversion efficiency (PCE) of 20.34% is achieved for doctor-bladed PSCs under the humidity over 80% with a device structure of ITO/SnO2 /MAPbI3 /Spiro-OMeTAD/Ag. It is the highest PCEs for one-step solution-processed MAPbI3 PSCs without antisolvent assistance. The research provides a facile and robust large-scale deposition technique to fabricate highly efficient and stable PSCs under a wide range of humidity, even with the humidity over 80%.

Toxicological effects and transcriptome mechanisms of rice (Oryza sativa L.) under stress of quinclorac and polystyrene nanoplastics.

The absorption and accumulation of nanoplastics (NPs) by plants is currently attracting considerable attention. NPs also tend to adsorb surrounding organic pollutants, such as pesticides, which can damage plants. However, molecular mechanisms underlying the phytotoxicity of NPs are not sufficiently researched. Therefore, we analyzed the toxicological effects of 50 mg/L polystyrene NPs (PS 50 nm) and 5 mg/L the herbicide quinolinic (QNC) on rice (Oryza sativa L.) using 7-day hydroponic experiments, explaining the corresponding mechanisms by transcriptome analysis. The main conclusion is that all treatments inhibit rice growth and activate the antioxidant level. Compared with CK, the inhibition rates of PS, QNC, and PS+QNC on rice shoot length were 3.95%, 6.68%, and 11.43%, respectively. The gene ontology (GO) term photosynthesis was significantly enriched by QNC, and the combination PS+QNC significantly enriched the GO terms of amino sugar and nucleotide sugar metabolisms. The chemicals QNC and PS+QNC significantly affected the Kyoto Encyclopedia of Genes and Genomes (KEGG) of the MAPK signaling pathway, plant hormone signal transduction, and plant-pathogen interaction. Our findings provide a new understanding of the phytotoxic mechanisms and environmental impacts of the interactions between NPs and pesticides. It also provides insights into the impact of NPs and pesticides on plants in the agricultural system.

Dual-modal radiomics for predicting cervical lymph node metastasis in papillary thyroid carcinoma.

BACKGROUND: Preoperative prediction of cervical lymph node metastasis (CLNM) in patients with papillary thyroid carcinoma (PTC) is significant for surgical decision-making. OBJECTIVE: This study aims to develop a dual-modal radiomics (DMR) model based on grayscale ultrasound (GSUS) and dual-energy computed tomography (DECT) for non-invasive CLNM in PTC. METHODS: In this study, 348 patients with pathologically confirmed PTC at Jiangsu University Affiliated People's Hospital who completed preoperative ultrasound (US) and DECT examinations were enrolled and randomly assigned to training (n = 261) and test (n = 87) cohorts. The enrolled patients were divided into two groups based on pathology findings namely, CLNM (n = 179) and CLNM-Free (n = 169). Radiomics features were extracted from GSUS images (464 features) and DECT images (960 features), respectively. Pearson correlation coefficient (PCC) and the least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation were then used to select CLNM-related features. Based on the selected features, GSUS, DECT, and GSUS combined DECT radiomics models were constructed by using a Support Vector Machine (SVM) classifier. RESULTS: Three predictive models based on GSUS, DECT, and a combination of GSUS and DECT, yielded performance of areas under the curve (AUC) = 0.700 [95% confidence interval (CI), 0.662-0.706], 0.721 [95% CI, 0.683-0.727], and 0.760 [95% CI, 0.728-0.762] in the training dataset, and AUC = 0.643 [95% CI, 0.582-0.734], 0.680 [95% CI, 0.623-0.772], and 0.744 [95% CI, 0.686-0.784] in the test dataset, respectively. It shows that the predictive model combined GSUS and DECT outperforms both models using GSUS and DECT only. CONCLUSIONS: The newly developed combined radiomics model could more accurately predict CLNM in PTC patients and aid in better surgical planning.

Comprehensive analysis of the prognosis and immune infiltration landscape of RNA methylation-related subtypes in pancreatic cancer.

BACKGROUND: RNA methylation refers to a form of methyl modification in RNA that modulates various epigenetic alterations. Mounting studies have focused on its potential mechanisms in cancer initiation and progression. However, the prognostic value and potential role of RNA methylation in the immune microenvironment of pancreatic cancer remain unclear. METHODS: Comprehensive bioinformatics analysis was performed to illuminate the expression profiles of RNA methylation modulators. In addition, the ConsensusClusterPlus algorithm was utilized to identify two remarkably different subtypes, and a feasible risk stratification method was established to accurately estimate prognosis. In addition, we validated our signature at the cytology and histology levels and conducted functional experiments to explore the biological functions of our key genes. RESULTS: Two subtypes with remarkable survival differences were identified by the consensus clustering algorithm. Cluster 2 tended to have higher expression levels of RNA methylation regulators and to be the high RNA methylation group. In addition, cluster 1 exhibited a significantly higher abundance of almost all immune cells and increased immune checkpoint expression compared to cluster 2. Chemotherapeutic sensitivity analysis indicated that there were significant differences in the sensitivity of four of the six drugs between different subgroups. Mutation investigation revealed a higher mutation burden and a higher number of mutations in cluster 2. An accurate and feasible risk stratification method was established based on the expression of key genes of each subtype. Patients with low risk scores exhibited longer survival times in one training (TCGA) and two validation cohorts (ICGC, GSE57495), with p values of 0.001, 0.0081, and 0.0042, respectively. In addition, our signature was further validated in a cohort from Fudan University Shanghai Cancer Center. The low-risk group exhibited higher immune cell abundance and immune checkpoint levels than the high-risk group. The characteristics of the low-risk group were consistent with those of cluster 1: higher stromal score, estimate score, and immune score and lower tumor purity. Additionally, cell function investigations suggested that knockdown of CDKN3 remarkably inhibited the proliferation and migration of pancreatic cancer cells. CONCLUSIONS: RNA methylation has a close correlation with prognosis, immune infiltration and therapy in pancreatic cancer. Our subtypes and risk stratification method can accurately predict prognosis and the efficacy of immune therapy and chemotherapy.

NAGNN: Classification of COVID-19 based on neighboring aware representation from deep graph neural network.

COVID-19 pneumonia started in December 2019 and caused large casualties and huge economic losses. In this study, we intended to develop a computer-aided diagnosis system based on artificial intelligence to automatically identify the COVID-19 in chest computed tomography images. We utilized transfer learning to obtain the image-level representation (ILR) based on the backbone deep convolutional neural network. Then, a novel neighboring aware representation (NAR) was proposed to exploit the neighboring relationships between the ILR vectors. To obtain the neighboring information in the feature space of the ILRs, an ILR graph was generated based on the k-nearest neighbors algorithm, in which the ILRs were linked with their k-nearest neighboring ILRs. Afterward, the NARs were computed by the fusion of the ILRs and the graph. On the basis of this representation, a novel end-to-end COVID-19 classification architecture called neighboring aware graph neural network (NAGNN) was proposed. The private and public data sets were used for evaluation in the experiments. Results revealed that our NAGNN outperformed all the 10 state-of-the-art methods in terms of generalization ability. Therefore, the proposed NAGNN is effective in detecting COVID-19, which can be used in clinical diagnosis.

Diagnosis of COVID-19 Pneumonia via a Novel Deep Learning Architecture.

COVID-19 is a contagious infection that has severe effects on the global economy and our daily life. Accurate diagnosis of COVID-19 is of importance for consultants, patients, and radiologists. In this study, we use the deep learning network AlexNet as the backbone, and enhance it with the following two aspects: 1) adding batch normalization to help accelerate the training, reducing the internal covariance shift; 2) replacing the fully connected layer in AlexNet with three classifiers: SNN, ELM, and RVFL. Therefore, we have three novel models from the deep COVID network (DC-Net) framework, which are named DC-Net-S, DC-Net-E, and DC-Net-R, respectively. After comparison, we find the proposed DC-Net-R achieves an average accuracy of 90.91% on a private dataset (available upon email request) comprising of 296 images while the specificity reaches 96.13%, and has the best performance among all three proposed classifiers. In addition, we show that our DC-Net-R also performs much better than other existing algorithms in the literature. Supplementary Information: The online version contains supplementary material available at 10.1007/s11390-020-0679-8.

ResGNet-C: A graph convolutional neural network for detection of COVID-19.

The widely spreading COVID-19 has caused thousands of hundreds of mortalities over the world in the past few months. Early diagnosis of the virus is of great significance for both of infected patients and doctors providing treatments. Chest Computerized tomography (CT) screening is one of the most straightforward techniques to detect pneumonia which was caused by the virus and thus to make the diagnosis. To facilitate the process of diagnosing COVID-19, we therefore developed a graph convolutional neural network ResGNet-C under ResGNet framework to automatically classify lung CT images into normal and confirmed pneumonia caused by COVID-19. In ResGNet-C, two by-products named NNet-C, ResNet101-C that showed high performance on detection of COVID-19 are simultaneously generated as well. Our best model ResGNet-C achieved an averaged accuracy at 0.9662 with an averaged sensitivity at 0.9733 and an averaged specificity at 0.9591 using five cross-validations on the dataset, which is comprised of 296 CT images. To our best knowledge, this is the first attempt at integrating graph knowledge into the COVID-19 classification task. Graphs are constructed according to the Euclidean distance between features extracted by our proposed ResNet101-C and then are encoded with the features to give the prediction results of CT images. Besides the high-performance system, which surpassed all state-of-the-art methods, our proposed graph construction method is simple, transferrable yet quite helpful for improving the performance of classifiers, as can be justified by the experimental results.

Optimized deposition time boosts the performance of Prussian blue modified nanoporous gold electrodes for hydrogen peroxide monitoring.

Prussian blue (PB) modified nanoporous gold (NPG) electrodes exhibit great potential for improving the detection sensitivity and stability for hydrogen peroxide monitoring. The NPG provides large surface-to-volume ratio as well as diffusion 'highways' to assist the transfer of the ions. In the present work, we optimized the deposition time for NPG fabrication and examine the electrochemical performance of the electrodes. A critical deposition time on the electrochemical performances including linear range, operational stability and sensitivity was experimentally determined. Below and above such a deposition time, two different growth patterns of the microstructures were observed. This transition of deposited structures corresponding to the critical time results in different pathways for electron transfer and ion diffusivity through PB lattice.

The rice pds1 locus genetically interacts with partner to cause panicle exsertion defects and ectopic tillers in spikelets.

BACKGROUND: Rice (Oryza sativa L.) is a staple food crop worldwide. Its yield and quality are affected by its tillering pattern and spikelet development. Although many genes involved in the vegetative and reproductive development of rice have been characterized in previous studies, the genetic mechanisms that control axillary tillering, spikelet development, and panicle exsertion remain incompletely understood. RESULTS: Here, we characterized a novel rice recombinant inbred line (RIL), panicle exsertion defect and aberrant spikelet (pds). It was derived from a cross between two indica varieties, S142 and 430. Intriguingly, no abnormal phenotypes were observed in the parents of pds. This RIL exhibited sheathed panicles at heading stage. Still, a small number of tillers in pds plants were fully exserted from the flag leaves. Elongated sterile lemmas and rudimentary glumes (occurred occasionally) were observed in the spikelets of the exserted panicles and were transformed into palea/lemma-like structures. Furthermore, more interestingly, tillers occasionally grew from the axils of the elongated rudimentary glumes. Via genetic linkage analysis, we found that the abnormal phenotype of pds manifesting as genetic incompatibility or hybrid weakness was caused by genetic interaction between a recessive locus, pds1, which was derived from S142 and mapped to chromosome 8, and a locus pds2, which not yet mapped from 430. We fine-mapped pds1 to an approximately 55-kb interval delimited by the markers pds-4 and 8 M3.51. Six RGAP-annotated ORFs were included in this genomic region. qPCR analysis revealed that Loc_Os080595 might be the target of pds1 locus, and G1 gene might be involved in the genetic mechanism underlying the pds phenotype. CONCLUSIONS: In this study, histological and genetic analyses revealed that the pyramided pds loci resulted in genetic incompatibility or hybrid weakness in rice might be caused by a genetic interaction between pds loci derived from different rice varieties. Further isolation of pds1 and its interactor pds2, would provide new insight into the molecular regulation of grass inflorescence development and exsertion, and the evolution history of the extant rice.

Comparative proteomic analysis of QTL CTS-12 derived from wild rice (Oryza rufipogon Griff.), in the regulation of cold acclimation and de-acclimation of rice (Oryza sativa L.) in response to severe chilling stress.

BACKGROUND: Rice (Oryza sativa L.) is a thermophilic crop vulnerable to chilling stress. However, common wild rice (Oryza rufipogon Griff.) in Guangxi (China) has the ability to tolerate chilling stress. To better understand the molecular mechanisms underlying chilling tolerance in wild rice, iTRAQ-based proteomic analysis was performed to examine CTS-12, a major chilling tolerance QTL derived from common wild rice, mediated chilling and recovery-induced differentially expressed proteins (DEPs) between the chilling-tolerant rice line DC90 and the chilling-sensitive 9311. RESULTS: Comparative analysis identified 206 and 155 DEPs in 9311 and DC90, respectively, in response to the whole period of chilling and recovery. These DEPs were clustered into 6 functional groups in 9311 and 4 in DC90. The majority were enriched in the 'structural constituent of ribosome', 'protein-chromophore linkage', and 'photosynthesis and light harvesting' categories. Short Time-series Expression Miner (STEM) analysis revealed distinct dynamic responses of both chloroplast photosynthetic and ribosomal proteins between 9311 and DC90. CONCLUSION: CTS-12 might mediate the dynamic response of chloroplast photosynthetic and ribosomal proteins in DC90 under chilling (cold acclimation) and recovery (de-acclimation) and thereby enhancing the chilling stress tolerance of this rice line. The identified DEPs and the involvement of CTS-12 in mediating the dynamic response of DC90 at the proteomic level illuminate and deepen the understanding of the mechanisms that underlie chilling stress tolerance in wild rice.

Triggered single photon emission up to 77K from ordered array of surface curvature-directed mesa-top GaAs/InGaAs single quantum dots.

We demonstrate triggered single photon emission up to 77K from an ordered 5x8 array of InGaAs single quantum dots (SQDs). The SQDs are grown selectively on patterned mesa tops utilizing substrate-encoded size-reducing epitaxy (SESRE). It exploits designed surface-curvature stress gradients to preferentially direct atom migration from mesa sidewalls to the top during growth. The emission from the SQDs exhibits a g(2)(0) of 0.19 ± 0.03 at 8K and decent emission spectral uniformity (standard deviation <1% of emission wavelength). The SESRE QDs are inherently compatible with on-chip integrated light manipulation elements, thereby enabling a path towards integrated nanophotonic systems for quantum information processing.

Graph reasoning method enhanced by relational transformers and knowledge distillation for drug-related side effect prediction.

Identifying the side effects related to drugs is beneficial for reducing the risk of drug development failure and saving the drug development cost. We proposed a graph reasoning method, RKDSP, to fuse the semantics of multiple connection relationships, the local knowledge within each meta-path, the global knowledge among multiple meta-paths, and the attributes of the drug and side effect node pairs. We constructed drug-side effect heterogeneous graphs consisting of the drugs, side effects, and their similarity and association connections. Multiple relational transformers were established to learn node features from diverse meta-path semantic perspectives. A knowledge distillation module was constructed to learn local and global knowledge of multiple meta-paths. Finally, an adaptive convolutional neural network-based strategy was presented to adaptively encode the attributes of each drug-side effect node pair. The experimental results demonstrated that RKDSP outperforms the compared state-of-the-art prediction approaches.

Impacts of Millipedes on Acari and Collembola Communities-A Microcosm Experiment.

Ecosystem engineers influence the structure and function of soil food webs through non-trophic interactions. The activity of large soil animals, such as earthworms, has a significant impact on the soil microarthropod community. However, the influence of millipedes on soil microarthropod communities remains largely unknown. In this microcosm experiment, we examined the effects of adding, removing, and restricting millipede activity on Acari and Collembola communities in litter and soil by conducting two destructive sampling sessions on days 10 and 30, respectively. At the time of the first sampling event (10 d), Acari and Collembola abundance was shown to increase and the alpha diversity went higher in the treatments with millipedes. At the time of the second sampling event (30 d), millipedes significantly reduced the Collembola abundance and alpha diversity. The results were even more pronounced as the millipedes moved through the soil, which caused the collembolans to be more inclined to inhabit the litter, which in turn resulted in the increase in the abundance and diversity of Acari in the soil. The rapid growth of Collembola in the absence of millipedes significantly inhibited the abundance of Acari. The presence of millipedes altered the community structure of Acari and Collembola, leading to a stronger correlation between the two communities. Changes in these communities were driven by the dominant taxa of Acari and Collembola. These findings suggest that millipedes, as key ecosystem engineers, have varying impacts on different soil microarthropods. This study enhances our understanding of biological interactions and offers a theoretical foundation for soil biodiversity conservation.

A nomogram for predicting the necessity of tracheostomy after severe acute brain injury in patients within the neurosurgery intensive care unit: A retrospective cohort study.

Objective: This retrospective study was aimed to develop a predictive model for assessing the necessity of tracheostomy (TT) in patients admitted to the neurosurgery intensive care unit (NSICU). Method: We analyzed data from 1626 NSICU patients with severe acute brain injury (SABI) who were admitted to the Department of NSICU at the Affiliated People's Hospital of Jiangsu University between January 2021 and December 2022. Data of the patients were retrospectively obtained from the clinical research data platform. The patients were randomly divided into training (70%) and testing (30%) cohorts. The least absolute shrinkage and selection operator (LASSO) regression identified the optimal predictive features. A multivariate logistic regression model was then constructed and represented by a nomogram. The efficacy of the model was evaluated based on discrimination, calibration, and clinical utility. Results: The model highlighted six predictive variables, including the duration of NSICU stay, neurosurgery, orotracheal intubation time, Glasgow Coma Scale (GCS) score, systolic pressure, and respiration rate. Receiver operating characteristic (ROC) analysis of the nomogram yielded area under the curve (AUC) values of 0.854 (95% confidence interval [CI]: 0.822-0.886) for the training cohort and 0.865 (95% CI: 0.817-0.913) for the testing cohort, suggesting commendable differential performance. The predictions closely aligned with actual observations in both cohorts. Decision curve analysis demonstrated that the numerical model offered a favorable net clinical benefit. Conclusion: We developed a novel predictive model to identify risk factors for TT in SABI patients within the NSICU. This model holds the potential to assist clinicians in making timely surgical decisions concerning TT.

Hypothermia protects the integrity of corticospinal tracts and alleviates mitochondria injury after intracerebral hemorrhage in mice.

Disruption of corticospinal tracts (CST) is a leading factor for motor impairments following intracerebral hemorrhage (ICH) in the striatum. Previous studies have shown that therapeutic hypothermia (HT) improves outcomes of ICH patients. However, whether HT has a direct protection effect on the CST integrity and the underlying mechanisms remain largely unknown. In this study, we employed a chemogenetics approach to selectively activate bilateral warm-sensitive neurons in the preoptic areas to induce a hypothermia-like state. We then assessed effects of HT treatment on the integrity of CST and motor functional recovery after ICH. Our results showed that HT treatment significantly alleviated axonal degeneration around the hematoma and the CST axons at remote midbrain region, ultimately promoted skilled motor function recovery. Anterograde and retrograde tracing revealed that HT treatment protected the integrity of the CST over an extended period. Mechanistically, HT treatment prevented mitochondrial swelling in degenerated axons around the hematoma, alleviated mitochondrial impairment by reducing mitochondrial ROS accumulation and improving mitochondrial membrane potential in primarily cultured cortical neurons with oxyhemoglobin treatment. Serving as a proof of principle, our study provided novel insights into the application of HT to improve functional recovery after ICH.

Inducing repetitive action potential firing in neurons via synthesized photoresponsive nanoscale cellular prostheses.

Recently we reported an analysis that examined the potential of synthesized photovoltaic functional abiotic nanosystems (PVFANs) to modulate membrane potential and activate action potential firing in neurons. Here we extend the analysis to delineate the requirements on the electronic energy levels and the attendant photophysical properties of the PVFANs to induce repetitive action potential under continuous light, a capability essential for the proposed potential application of PVFANs as retinal cellular prostheses to compensate for loss of photoreceptors. We find that repetitive action potential firing demands two basic characteristics in the electronic response of the PVFANs: an exponential dependence of the PVFAN excited state decay rate on the membrane potential and a three-state system such that, following photon absorption, the electron decay from the excited state to the ground state is via intermediate state(s) whose lifetime is comparable to the refractory time following an action potential. FROM THE CLINICAL EDITOR: In this study, the potential of synthetic photovoltaic functional abiotic nanosystems (PVFANs) is examined under continuous light to modulate membrane potential and activate action potential firing in neurons with the proposed potential application of PVFANs as retinal cellular prostheses.

Diagnosis of cerebral microbleed via VGG and extreme learning machine trained by Gaussian map bat algorithm.

Cerebral microbleed (CMB) is a serious public health concern. It is associated with dementia, which can be detected with brain magnetic resonance image (MRI). CMBs often appear as tiny round dots on MRIs, and they can be spotted anywhere over brain. Therefore, manual inspection is tedious and lengthy, and the results are often short in reproducible. In this paper, a novel automatic CMB diagnosis method was proposed based on deep learning and optimization algorithms, which used the brain MRI as the input and output the diagnosis results as CMB and non-CMB. Firstly, sliding window processing was employed to generate the dataset from brain MRIs. Then, a pre-trained VGG was employed to obtain the image features from the dataset. Finally, an ELM was trained by Gaussian-map bat algorithm (GBA) for identification. Results showed that the proposed method VGG-ELM-GBA provided better generalization performance than several state-of-the-art approaches.

Efficient Layer-Wise N:M Sparse CNN Accelerator with Flexible SPEC: Sparse Processing Element Clusters.

Recently, the layer-wise N:M fine-grained sparse neural network algorithm (i.e., every M-weights contains N non-zero values) has attracted tremendous attention, as it can effectively reduce the computational complexity with negligible accuracy loss. However, the speed-up potential of this algorithm will not be fully exploited if the right hardware support is lacking. In this work, we design an efficient accelerator for the N:M sparse convolutional neural networks (CNNs) with layer-wise sparse patterns. First, we analyze the performances of different processing element (PE) structures and extensions to construct the flexible PE architecture. Second, the variable sparse convolutional dimensions and sparse ratios are involved in the hardware design. With a sparse PE cluster (SPEC) design, the hardware can efficiently accelerate CNNs with the layer-wise N:M pattern. Finally, we employ the proposed SPEC into the CNN accelerator with flexible network-on-chip and specially designed dataflow. We implement hardware accelerators on Xilinx ZCU102 FPGA and Xilinx VCU118 FPGA and evaluate them with classical CNNs such as Alexnet, VGG-16, and ResNet-50. Compared with existing accelerators designed for structured and unstructured pruned networks, our design achieves the best performance in terms of power efficiency.