Fine Network Research Articles

Fine-Graine Visual Classification with Aggregated Object Localization and Salient Feature Suppression

Fine-grained visual classification (FGVC) is desired to classify sub-classes of objects in the same super-class. For the FGVC tasks, it is necessary to find subtle yet discriminative information from local areas. However, traditional FGVC approaches tended to extract strong discriminative features, and overlook some subtle yet useful features. Besides, current methods ignore the influence of background noises on feature extraction. Therefore, aggregated object localization combined with salient feature suppression are proposed, which is a stacked network. First, the feature maps extracted by the coarse network are fed into aggregated object localization to obtain complete foreground object in an image. Secondly, the refined features obtained through zooming in complete foreground object are fed into fine network. Finally, through finer network processing, the feature maps are fed into salient feature suppression module to find more valuable region discriminative features for classification. Experiment results on two datasets show that our proposed method can get superior result compared with state-of-the-art methods.

A Coarse-to-Fine Deep Generative Model With Spatial Semantic Attention for High-Resolution Remote Sensing Image Inpainting

Large area array CCD splicing technology and hardware damage lead to the lack of information in high-resolution remote sensing images (HRSIs), which seriously restricts the application of remote sensing images. However, the complex spatial geographical relationship and the serious spatial heterogeneity will cause texture blur and incomplete structure, especially in the boundary and highly textured regions of HRSI. To deal with the challenge, this article proposes a coarse-to-fine deep generative model with a novel spatial semantic attention (SSA) mechanism. SSA is deployed in the encoder and decoder of the network to ensure the continuity of local features and the relevance of global semantic information and embedded in fine network. In the coarse network, a multilevel loss is proposed to provide more accurate information. Fine network performs more detailed image inpainting operations based on coarse network. Extensive experiments on three datasets, including GF2 satellite images, aerial images, and natural ones, demonstrate that our proposed model has better inpainting effect than the state-of-the-art algorithms.

Moyamoya Disease Associated with Hyperhomocysteinemia: A Rare Cause of Stroke in Sub-Saharan Africa

Introduction: Moyamoya disease (MMD) is a chronic cerebrovascular disorder characterized by progressive bilateral occlusion of the supraclinoid internal carotid artery (ICA) and its main branches, associated with the development of fine collateral networks, especially adjacent to the site of occlusion in the deep areas of the brain. MMD frequently occurs in East Asian populations, including pediatric and adult patients, and may lead to ischemic or hemorrhagic stroke, headache, epilepsy or transient ischemic attack. The majority is presumed to be of genetic origin and few cases of thrombophilia have been reported in MMD. We report a case of MMD in a young sub-Saharan African woman associated with Hyperhomocysteninemia (HHCys). Observation: A 33-year-old female was seen for aphasia, which had suddenly appeared fourteen days before, and was associated with memory impairment. She mentioned chronic headaches during the 10 previous years, of frontal seat, without migraine criteria. She had no history of hypertension, no obesity, no known dyslipidemia, and there was no family history of stroke. Neurological examination noted Wernicke’s aphasia, with discreet right central facial palsy, associated with memory impairment. The brain MRI revealed semi recent infarct in the left parietal lobe, sequelae in the left temporo-occipital and right frontal with lacunar ischemic lesions, associated with several foci of sub- cortical demyelination. MRA showed stenosis of the supraclinoid carotid arteries and their division branches. In addition, there was a stenosis of the posterior cerebral arteries, with a network of anastomosis at the level of the base of the skull. An investigation of thrombophilia showed hyperhomocysteinemia (HHCys). The diagnosis of moyamoya disease associated with hyperhomocysteinemia was mentioned. She was treated by Aspirin combined to folic acid. After a 3-year setback the patient was asymptomatic. Conclusion: MMD is a rare cause of stroke in sub-Saharan Africa. MRA should be performed whenever possible, as well as investigation for thrombophilia when it’s available. Combined antiplatelet agent with folic acid for ischemic stroke prevention is effective.

Quantum-Enhanced Deep Learning-Based Lithology Interpretation From Well Logs

Lithology interpretation is important for understanding subsurface properties. Yet, the common manual well log interpretation is usually with low efficiency and bad consistency. Therefore, the automatic well log interpretation tools based on machine learning and deep learning have been developed. Although the state-of-the-art sophisticated models can show fine interpretation performance with acceptable accuracies, “blind” tests do not always exhibit satisfactory results because of the complexity of lithology interpretation with respect to subsurface rock properties and the data-labeling quality. To solve this generalization challenge, we propose to leverage the parameterized quantum circuits in the deep-learning model. The quantum computing takes advantages of the superposition and entanglement quantum systems, which could potentially endow the generalization power or capability to the deep-learning model. Using the proposed quantum-enhanced deep-learning (QEDL) model, we have tested the model performance on field well log data from different wells. Compared with the classic fine convolutional neural network (CNN) model and the long short-term memory (LSTM) model, the proposed QEDL model achieves comparable model performance with a clearly improved generalization power for interpreting both thin and thick lithology layers. In addition, because of the quantum circuit structure, the QEDL model needs much fewer model parameters than LSTM and CNN models, i.e., the QEDL parameter number in our study can be approximately 75% less than that of LSTM and 89% less than that of CNN.

In-situ formation of particle reinforced Aluminium matrix composites by laser powder bed fusion of Fe2O3/AlSi12 powder mixture using laser melting/remelting strategy

In-situ preparation of particle reinforced Al matrix composites (PRAMCs) by laser powder bed fusion (LPBF) is a promising strategy to strengthen Al-based alloys. The laser-driven thermite reaction can be a practical mechanism to in-situ synthesise PRAMCs. However, the introduction of elements oxygen by adding Fe2O3 makes the powder mixture highly sensitive to form porosity and Al2O3 film during LPBF, bringing challenges to prepare dense materials. This work develops an LPBF processing strategy combined with consecutive high-energy laser melting scanning and low-energy laser remelting scanning to prepare dense PRAMCs from Fe2O3/AlSi12 powder mixture. A high relative density (98.2 ± 0.55 %) was successfully obtained by optimising laser melting (Emelting) and remelting energy density (Eremelting) to Emelting = 35 J/mm2 and Eremelting = 5 J/mm2. Results reveal the necessity to increase Emelting to improve metal liquid’s spreading/wetting by breaking up Al2O3 films surrounding molten pools; however, the high-energy laser melting produced much porosity. Low-energy laser remelting could close the resulting internal pores, backfill open gaps and smoothen solidified surfaces. Although with two-times laser scanning, the microstructure still shows fine cellular Si networks with Al grains inside (grain size 370 nm) and in-situ nano-precipitates (Al2O3, Si and Al-Fe(-Si) intermetallics). Finally, the fine microstructure, nano-structured dispersion strengthening and high-level densification strengthen the prepared in-situ PRAMCs, reaching yield strength of 426 ± 4 MPa and tensile strength of 473 ± 6 MPa. Furthermore, the results can provide valuable information to process other powder mixtures with severe porosity/oxide-film formation potential considering the evidenced contribution of laser melting/remelting strategy to densify material and obtain good mechanical properties during LPBF.

Confocal Microscopy of Filtering Blebs after Trabeculectomy.

Filtration surgery is the most effective method of lowering intraocular pressure (IOP) in patients with insufficient medical control. It consists in facilitating the drainage of the intraocular fluid (IOF) from the anterior chamber to the subconjunctival space and subsequent lowering of IOP. The formation of filtration blebs (FB) and the processes of scarring occurring in the conjunctiva are of particular importance in glaucoma surgery. In many cases, the appearance of FB does not match the IOP values, and what causes the failure after trabeculectomy often remains unclear. Often, over time, there is a change in the structure of the FB, as fibrous tissue grows, which prevents the IOF drainage. Laser scanning in vivo confocal microscopy is a non-invasive study allowing the production of layered images at the microstructural level with high resolution of both the cornea and other structures of the anterior ocular surface. To evaluate the morphological structure and function of filtering blebs after trabeculectomy using in vivo confocal microscopy taking into account the type of implant and when the surgery was performed. The study included 33 patients, 46 eyes with glaucoma. Twenty-six of the eyes had primary open-angle glaucoma (POAG), 18 eyes had pseudoexfoliative glaucoma and 2 eyes had juvenile glaucoma. All patients underwent trabeculectomy with fornix-based flap, and three of the eyes underwent retrabeculectomy. Mitomicyn C (MMC) was administered intraoperatively to all patients. The study of the filtering bleb was performed by in vivo confocal microscopy (CFM) (Heidelberg Retina Tomograph II (HRT II) /Rostock Cornea Module/ (Heidelberg Engineering GmbH, Heidelberg, Germany), the period from trabeculectomy and examination being from 1 year to 22 years. An Express implant was placed in 14 eyes, Ologen implant in 7 eyes, and 25 eyes had no implant placed. In the analysis of the morphological structure of the filtering blebs, three indicators were evaluated: the type of epithelium, the type of stroma, and blood vessels. Statistical significance was established with regard to the function and morphological structure of the filtering bleb (p=0.009). Blebs with fine collagen mesh and dense collagen mesh demonstrate good function. In the case of blebs with insufficient function, those with a dense collagen network and hyper-reflective tissue predominated and there were no blebs with a fine collagen network, and in non-functioning blebs most common were those with a pronounced collagen network and hyper-reflective tissue. With regard to vascularization, we found that the functioning blebs in the shortest postoperative period were dominated by those with one blood vessel (stage 1) and there was no stage 3, with weak tortuosity, while in non-functioning blebs in the late postoperative period, there was moderate to severe vascularization and tortuosity (p=0.037), (p=0.043), (p=0.047), (p=0.021). The type of implant affects the tortuosity of the blood vessels of the filtering bleb (p=0.026). The blebs with Express implants show a slight tortuosity, followed by the blebs with Ologen implants. The highest percentage of highly kinked blood vessels occurred in blebs without an implant. In vivo confocal microscopy is an innovative method which allows visualization of the internal structure of the filtering blebs at a cellular level, giving us a new insight into the ongoing healing processes, premising the function of the filtering blebs after glaucoma surgery.

Renal tumors segmentation in abdomen CT Images using 3D-CNN and ConvLSTM

Renal tumor is one of the common tumors with high incidence, and accurate segmentation of renal tumors is helpful for preoperative evaluation. Computed Tomography (CT) plays an important role in the treatment of renal tumors and accurate segmentation of tumors in CT images may provide critical information for surgery. In this paper, a segmentation approach based on deep learning with limited computation cost is proposed to improve the segmentation accuracy for kidneys and renal tumors. Firstly, a pre-trained restruction network is presented to alleviate small samples problems, which utilizes abdominal CT data to transfer network model effectively; Then, prior contour-assisted channel is introduced in two-dimensional network to segment the region of interest which contains kidneys and renal tumors and act as the input of the subsequente fine segmentation network; Finally, convolutional long short-term memory (ConvLSTM) is employed to extract spatial correlation information between slices and combined with a three-dimensional convolutional neural networks for fine segmentation. Several experiments on the 2019 renal tumor segmentation challenge(Kits19) dataset are designed to evaluate the performance of the proposed method, and the mean segmentation accuracy for kidneys and renal tumors are 96.39% and 78.91% for cross validation tests, which outperforms the other neural network algorithms, including 3D Res-Unet with 95.4% and 72.35%.

Temperature-induced response in algal cell surface properties and behaviour: an experimental approach

Microalgae are considered an accurate indicator of ecosystem perturbations induced by global climate change. The present work aims to investigate the alteration of temperature on surface properties and behaviour of three algal species using the complementary surface methods (electrochemical and atomic force microscopy). The results showed that the temperature-induced response of algae is species-specific due to the structural features of the cell envelope. Wall-less algae experience the largest nanomechanical and chemical change, while algae with silicified walls show the pronounced chemical change in the degree of hydrophobicity. Alterations of surface properties suggest a molecular modification of the algal barrier and cytoskeletal rearrangements due to a change in cell size, while algal morphology reveals no change. The physiological activity of cells showed a different organisation of released extracellular substances in the form of fine fibrillar structures, aggregated particles, and dense networks. Both types of algal responses, physiological activity, and molecular modification of the cell barrier determine the cell adhesion and motility. This study highlights the role of surface properties in cell-substrate and cell–cell interactions, which is important for the understanding of algal behaviour at natural interfaces and the mechanism of algal biofilm and aggregate formation in aquatic systems under the stress.Graphical abstract

Fully automated multiorgan segmentation of female pelvic magnetic resonance images with coarse-to-fine convolutional neural network.

Brachytherapy combined with external beam radiotherapy (EBRT) is the standard treatment for cervical cancer and has been shown to improve overall survival rates compared to EBRT only. Magnetic resonance (MR) imaging is used for radiotherapy (RT) planning and image guidance due to its excellent soft tissue image contrast. Rapid and accurate segmentation of organs at risk (OAR) is a crucial step in MR image-guided RT. In this paper, we propose a fully automated two-step convolutional neural network (CNN) approach to delineate multiple OARs from T2-weighted (T2W) MR images. We employ a coarse-to-fine segmentation strategy. The coarse segmentation step first identifies the approximate boundary of each organ of interest and crops the MR volume around the centroid of organ-specific region of interest (ROI). The cropped ROI volumes are then fed to organ-specific fine segmentation networks to produce detailed segmentation of each organ. A three-dimensional (3-D) U-Net is trained to perform the coarse segmentation. For the fine segmentation, a 3-D Dense U-Net is employed in which a modified 3-D dense block is incorporated into the 3-D U-Net-like network to acquire inter and intra-slice features and improve information flow while reducing computational complexity. Two sets of T2W MR images (221 cases for MR1 and 62 for MR2) were taken with slightly different imaging parameters and used for our network training and test. The network was first trained on MR1 which was a larger sample set. The trained model was then transferred to the MR2 domain via a fine-tuning approach. Active learning strategy was utilized for selecting the most valuable data from MR2 to be included in the adaptation via transfer learning. The proposed method was tested on 20 MR1 and 32 MR2 test sets. Mean± SD dice similarity coefficients are 0.93± 0.04, 0.87± 0.03, and 0.80± 0.10 on MR1 and 0.94± 0.05, 0.88± 0.04, and 0.80± 0.05 on MR2 for bladder, rectum, and sigmoid, respectively. Hausdorff distances (95th percentile) are 4.18± 0.52, 2.54± 0.41, and 5.03± 1.31mm on MR1 and 2.89± 0.33, 2.24± 0.40, and 3.28± 1.08mm on MR2, respectively. The performance of our method is superior to other state-of-the-art segmentation methods. We proposed a two-step CNN approach for fully automated segmentation of female pelvic MR bladder, rectum, and sigmoid from T2W MR volume. Our experimental results demonstrate that the developed method is accurate, fast, and reproducible, and outperforms alternative state-of-the-art methods for OAR segmentation significantly (p<0.05).

Grid-based spatiotemporal modeling of ambient ozone to assess human exposure using environmental big data

Ozone (O3) pollution in China is increasing. It is the primary pollutant in summer and ambient O3 can lead to serious health problems for the public. Therefore, characterizing the spatiotemporal distribution of O3 is required for better environmental management and human exposure assessment. Statistical models, especially those based on machine learning, can be more convenient to use than chemical transport models and have shown improved accuracy. However, the quality of data affects model precision especially at fine spatiotemporal scales. Web-based environmental data can improve the spatial and temporal resolution of modeling data. This study applied high spatiotemporal resolution source information and emission inventories based on point of interest and real-time traffic data in a fine scale grid network to predict the O3 concentrations and assess the human exposure within Chengdu, China. The results showed that the web-based environmental data could be combined with statistical models such as random forest in air quality modeling. The model precision was high, especially at finer spatiotemporal scales. The R2 of the hourly and daily maximum 8-h mean concentrations of O3 models built in this study were 0.83 and 0.91 for sample-based cross-validation, and 0.79 and 0.90 for site-based cross-validation, respectively. Meteorological variables had the greatest impact on O3 concentrations especially sea-level pressure, temperature, vapor pressure, and humidity. People within the research area had a relatively high exposure level to pollutants over a longer time scale in the summer and spring. Findings from this work provide a good reference for related research on modeling air quality, and human health risk assessment using environmental big data.

Structural evolution of contact parts of the friction stir processing heat-resistant nickel alloy tool used for multi-pass processing of Ti6Al4V/(Cu+Al) system

Interaction between nickel base superalloy ZhS6U used as a friction processing tool for in-situ synthesis of intermetallics in Ti6Al4V and this titanium alloy metal has been studied. Severe abrasion-adhesion wear was localized on the tool’ pin areas while shoulder was protected by mechanically mixed layer. Generation of titanium enriched mechanically mixed layers (MML) on the tool's surface was observed whose chemical and phase compositions corresponded to those of Ti2Ni and Ni4Ti3 intermetallic compounds. Fine-crystalline transition layer (TL) was found below the MML which contained high amount of tungsten in the form of fine MC carbide network. Dissolution of initial M6C carbides and diffusion of their components to the MML where new MC-type carbide network formed after stopping the FSP. The MML fragments detached from the tool surface and then were intermixed with the titanium alloy stir zone.

Review: Pelvic nerves - from anatomy and physiology to clinical applications.

A prerequisite for nerve-sparing pelvic surgery is a thorough understanding of the topographic anatomy of the fine and intricate pelvic nerve networks, and their connections to the central nervous system. Insights into the functions of pelvic nerves will help to interpret disease symptoms correctly and improve treatment. In this article, we review the anatomy and physiology of autonomic pelvic nerves, including their topography and putative functions. The aim is to achieve a better understanding of the mechanisms of pelvic pain and functional disorders, as well as improve their diagnosis and treatment. The information will also serve as a basis for counseling patients with chronic illnesses. A profound understanding of pelvic neuroanatomy will permit complex surgery in the pelvis without relevant nerve injury.

Bi-SANet—Bilateral Network with Scale Attention for Retinal Vessel Segmentation

The segmentation of retinal vessels is critical for the diagnosis of some fundus diseases. Retinal vessel segmentation requires abundant spatial information and receptive fields with different sizes while existing methods usually sacrifice spatial resolution to achieve real-time reasoning speed, resulting in inadequate vessel segmentation of low-contrast regions and weak anti-noise interference ability. The asymmetry of capillaries in fundus images also increases the difficulty of segmentation. In this paper, we proposed a two-branch network based on multi-scale attention to alleviate the above problem. First, a coarse network with multi-scale U-Net as the backbone is designed to capture more semantic information and to generate high-resolution features. A multi-scale attention module is used to obtain enough receptive fields. The other branch is a fine network, which uses the residual block of a small convolution kernel to make up for the deficiency of spatial information. Finally, we use the feature fusion module to aggregate the information of the coarse and fine networks. The experiments were performed on the DRIVE, CHASE, and STARE datasets. Respectively, the accuracy reached 96.93%, 97.58%, and 97.70%. The specificity reached 97.72%, 98.52%, and 98.94%. The F-measure reached 83.82%, 81.39%, and 84.36%. Experimental results show that compared with some state-of-art methods such as Sine-Net, SA-Net, our proposed method has better performance on three datasets.

VIRTUAL RESTORATION OF MISSING PAINT LOSS OF MURAL BASED ON GENERATIVE ADVERSARIAL NETWORK

Abstract. Mural painting is one of the important cultural heritage reflecting the historical migration of the nation. In order to inherit these precious historical and cultural heritage, how to non - destructively and digitally protect and restore the existing murals has become an urgent task. The use of computer - assisted restoration of murals can not only save manpower and material resources, but also avoid secondary damage to the murals.However, most of the existing computer-assisted mural restoration algorithms use similar blocks with priority calculations and matching blocks in adjacent areas to guide mural restoration. There are some problems such as incoherent overall semantic structure, unnatural detail texture and inability to effectively repair large area missing remain to be solved. Aiming at the problems existing in the restoration of large area diseases such as paint loss and color fading in murals, we constructed a fine image restoration network model which based on generative adversarial network. A multi-scale dense matching repair network based on a generative adversarial network is constructed. First, the dense combination of dilated convolutions is used to improve the repair effect of detailed textures, Then, mean absolute Error, (Visual Geometry Group, VGG) feature matching, auto-guided regression, and geometric alignment are used as the loss function to guide the training of the generative network. Second, the discriminator with local and global branches is used to train the discriminant network, so that the repaired image is in the local and global content. Experiments were performed on the three mural data sets one by one. The results show that the network model can effectively restore the lines and faces in the murals. The images restored are not only coherent in semantic details, but also natural in color, which is conducive to the appreciation and display of murals. Thus, as one of the important directions of cultural heritage digital protection，the use of generative adversarial network in the digital restoration of ancient murals have been proved to be effective. It not only provides a reference for the true restoration of the murals but also means a lot to the preservation of murals.

A simple and green approach to the preparation of super tough IIR/SWCNTs nanocomposites with tunable and strain responsive electrical conductivity

Abstract Nanocomposites of single-wall carbon nanotubes in isobutylene isoprene rubber (IIR/SWCNTs) were successfully prepared by a simple and green wet process. The traditional melt mixing process and organic solvent dissolution suffered from unable to effectively disperse the SWCNTs of tangled structure, and degradation of polymer molecules, respectively. Our process very well avoided these two problems. The SWCNTs aqueous solutions emulsified by polyoxyethylene octyl phenol ether (OP-10) were firstly mixed and compounded with IIR rubber at a relatively high temperature, followed by the second step of melt compounding process with the addition of cross-linking agent and accelerators. The SWCNTs were dispersed uniformly, and a fine network was constructed in the matrix of the obtained IIR/SWCNTs nanocomposite with a low percolation threshold. With the concentration of SWCNTs as low as 2 phr, the IIR/SWCNTs nanocomposite received an electrical conductivity of 10−6∼10−3 S/cm, and a 71% improvement of tensile strength. By varying the loadings of SWCNTs in a certain range, the tensile strength, electrical conductivity, and dielectric property were found tunable. Besides, the nanocomposites also presented strain responsive specific resistance, excellent elongation (600–740%), and better heat resistance.

Keratin 7 Is a Constituent of the Keratin Network in Mouse Pancreatic Islets and Is Upregulated in Experimental Diabetes.

Keratin (K) 7 is an intermediate filament protein expressed in ducts and glands of simple epithelial organs and in urothelial tissues. In the pancreas, K7 is expressed in exocrine ducts, and apico-laterally in acinar cells. Here, we report K7 expression with K8 and K18 in the endocrine islets of Langerhans in mice. K7 filament formation in islet and MIN6 β-cells is dependent on the presence and levels of K18. K18-knockout (K18‒/‒) mice have undetectable islet K7 and K8 proteins, while K7 and K18 are downregulated in K8‒/‒ islets. K7, akin to F-actin, is concentrated at the apical vertex of β-cells in wild-type mice and along the lateral membrane, in addition to forming a fine cytoplasmic network. In K8‒/‒ β-cells, apical K7 remains, but lateral keratin bundles are displaced and cytoplasmic filaments are scarce. Islet K7, rather than K8, is increased in K18 over-expressing mice and the K18-R90C mutation disrupts K7 filaments in mouse β-cells and in MIN6 cells. Notably, islet K7 filament networks significantly increase and expand in the perinuclear regions when examined in the streptozotocin diabetes model. Hence, K7 represents a significant component of the murine islet keratin network and becomes markedly upregulated during experimental diabetes.

Coarse-to-Fine Siamese Network for Tooth Object Tracking

At present, object tracking algorithms based on deep learning have been widely used in the computer vision industry and has great application value. In this paper, we track the tooth, which has the problem of poor texture and high similarity, with the object tracking algorithm. In general, previously existing single object tracking algorithms are derived from the siamese network architecture. These algorithms have the problem of losing a tracked object with poor textures and high similarities. In order to address the issue of object loss, we put forward a novel coarse-to-fine siamese network (CFSN) for tooth object tracking. The coarse detection network is based on the features extracted by the Feature Pyramid Networks (FPN) and the Region Proposal Network (RPN). The fine detection network has a similar structure to the coarse detection network, and the main purpose is to fine-tune the results of the previous stage. Based on the coarse-to-fine network, a framework of multi-task is proposed, that is, the mask prediction branch is added to the fine detection network, which lead to significant improvements in tracking efficiency. Through the comprehensive experimental evaluation of the tooth dataset, we find that the CFSN model is more effective in tracking tooth objects than the existing methods.

Handwritten Bangla Numerical Digit Recognition Using Fine Regulated Deep Neural Network

The recognition of handwritten Bangla digit is providing significant progress on optical character recognition (OCR). It is a very critical task due to the similar pattern and alignment of handwriting digits. With the progress of modern research on optical character recognition, it is reducing the complexity of the classification task by several methods, a few problems encounter during recognition and wait to be solved with simpler methods. The modern emerging field of artificial intelligence is the Deep Neural Network, which promises a solid solution to these few handwritten recognition problems. This paper proposed a fine regulated deep neural network (FRDNN) for the handwritten numeric character recognition problem that uses convolutional neural network (CNN) models with regularization parameters which makes the model generalized by preventing the overfitting. This paper applied Traditional Deep Neural Network (TDNN) and Fine regulated deep neural network (FRDNN) models with a similar layer experienced on BanglaLekha-Isolated databases and the classification accuracies for the two models were 96.25% and 96.99%, respectively over 100 epochs. The network performance of the FRDNN model on the BanglaLekha-Isolated digit dataset was more robust and accurate than the TDNN model and depend on experimentation. Our proposed method is obtained a good recognition accuracy compared with other existing available methods.

The role of shot peening on liquation cracking in laser cladding of K447A nickel superalloy powders over its non-weldable cast structure

A novel strategy with shot peening (SP) pretreatment was proposed to prevent heat-affected zone liquation cracking during laser cladding of K447A nickel-based superalloy powder over its non-weldable cast structure. High dislocation density accumulated by the SP pretreatment drove the high-temperature zone of the heat-affected zone to recrystallize during the laser cladding process, as a result, the coarse columnar grain with an average dendrite stem and secondary dendrite arm width of 102.67 ± 4.24 μm was transformed into a fine equiaxed grain with an average size of 10 μm. SP could effectively inhibit liquation cracking. With the increase of SP duration, the tendency of the liquation cracking reduced. Compared with no SP BM, the total length of the liquation cracking in the longitudinal section of the laser cladding sample reduced from 1649 μm to 248.8 μm. In-situ experimental observation and process simulation were opted to study the evolution behavior of recrystallization. Recrystallization firstly developed in the region of the intergranular zone at ~1214 °C and gradually the base metal recrystallized between 1232 and 1242 °C except for the region with γ/γ′ eutectic structure. As the temperature rose further, a fine liquid film network developed in the range of 1260–1267 °C. Finally, the stress and strain-based criteria were exercised to evaluate the crack susceptibility of the liquid film. The fine liquid film network adequately constrained the liquation cracking by reducing the driving stress applied to the liquid film as well as by lowering the pressure drop of the liquid film caused by driving strain.

A Single-Arm, Open-Label, Long-Term Efficacy and Safety Study of Subcutaneous (SC) Romiplostim in Children with Immune Thrombocytopenia (ITP)

A Single-Arm, Open-Label, Long-Term Efficacy and Safety Study of Subcutaneous (SC) Romiplostim in Children with Immune Thrombocytopenia (ITP)