Search

Recent research has shown that deep learning models are vulnerable to adversarial attacks, including gradient attacks, which can lead to incorrect outputs. The existing gradient attack methods typically rely on repetitive multistep strategies to improve their attack success rates, resulting in longer training times and severe overfitting. To address these issues, we propose an adaptive perturbation-based gradient attack method with dual-loss optimization (APDL). This method adaptively adjusts the single-step perturbation magnitude based on an exponential distance function, thereby accelerating the convergence process. APDL achieves convergence in fewer than 10 iterations, outperforming the traditional nonadaptive methods and achieving a high attack success rate with fewer iterations. Furthermore, to increase the transferability of gradient attacks such as APDL across different models and reduce the effects of overfitting on the training model, we introduce a triple-differential logit fusion (TDLF) method grounded in knowledge distillation principles. This approach mitigates the edge effects associated with gradient attacks by adjusting the hardness and softness of labels. Experiments conducted on ImageNet-compatible datasets demonstrate that APDL is significantly faster than the commonly used nonadaptive methods, whereas the TDLF method exhibits strong transferability. [ABSTRACT FROM AUTHOR]

Two-photon microscopy is indispensable in cell and molecular biology for its high-resolution visualization of cellular and molecular dynamics. However, the inevitable low signal-to-noise conditions significantly degrade image quality, obscuring essential details and complicating morphological analysis. While existing denoising methods such as CNNs, Noise2Noise, and DeepCAD serve broad applications in imaging, they still have limitations in preserving texture structures and fine details in two-photon microscopic images affected by complex noise, particularly in sophisticated structures like neuronal synapses. To improve two-photon microscopy image denoising effectiveness, by experimenting on real two-photon microscopy images, we propose a novel deep learning framework, the UNet-Att model, which integrates a specifically tailored UNet++ architecture with attention mechanisms. Specifically, this approach consists of a sophisticated downsampling module for extracting hierarchical features at varied scales, and an innovative attention module that strategically emphasizes salient features during the integration process. The architecture is completed by an ingenious upsampling pathway that reconstructs the image with high fidelity, ensuring the retention of textural integrity. Additionally, the model supports end-to-end training, optimizing its denoising efficacy. The UNet-Att model proves to surpass mainstream algorithms in the dual objectives of denoising and preserving the textural intricacies of original images, which is evidenced by an increase of 9.42 dB in the high Peak Signal-to-Noise Ratio (PSNR) coupled with an improvement of 0.1131 in the Structural Similarity Index Measurement (SSIM). The ablation experiments reveal the effectiveness of each module. The associated Python packages and datasets of UNet-Att are freely available at . [ABSTRACT FROM AUTHOR]

Purpose: This study aims to design an auxiliary segmentation model for thyroid nodules to increase diagnostic accuracy and efficiency, thereby reducing the workload of medical personnel. Methods: This study proposes a Dual-Path Attention Mechanism (DPAM)-UNet++ model, which can automatically segment thyroid nodules in ultrasound images. Specifically, the model incorporates dual-path attention modules into the skip connections of the UNet++ network to capture global contextual information in feature maps. The model's performance was evaluated using Intersection over Union (IoU), F1_score, accuracy, etc. Additionally, a new integrated loss function was designed for the DPAM-UNet++ network. Results: Comparative experiments with classical segmentation models revealed that the DPAM-UNet++ model achieved an IoU of 0.7451, an F1_score of 0.8310, an accuracy of 0.9718, a precision of 0.8443, a recall of 0.8702, an Area Under Curve (AUC) of 0.9213, and an HD95 of 35.31. Except for the precision metric, this model outperformed the other models on all the indicators and achieved a segmentation effect that was more similar to that of the ground truth labels. Additionally, ablation experiments verified the effectiveness and necessity of the dual-path attention mechanism and the integrated loss function. Conclusion: The segmentation model proposed in this study can effectively capture global contextual information in ultrasound images and accurately identify the locations of nodule areas. The model yields excellent segmentation results, especially for small and multiple nodules. Additionally, the integrated loss function improves the segmentation of nodule edges, enhancing the model's accuracy in segmenting edge details. [ABSTRACT FROM AUTHOR]

Multiplexing information in light's fundamental attributes to create supplementary orthogonal data channels has been well heralded as an effective means for optical data storage with greatly enhanced capacities. However, robust decoding methods against inevitable crosstalks associated with experimental noise and writing imperfections as the increase of multiplexing dimensions represent a major hurdle preventing the effective practice of multi-dimensional optical recording. Here, we propose a deep-learning-based retrieval approach for robust decoding multiplexed information. An artificial neural network is trained to learn the crosstalks from multiplexed recording in disordered gold nanorod aggregates with loosened orthogonality constraints. The acquired raw readout images are analyzed by the trained neural network, which allows quick, high-fidelity, and reliable information retrieval from polarization-, wavelength-, and 3D spatially multiplexed data. The smart decoding protocol paves the way toward the mass-production ready and wide-spread application of high-capacity multi-dimensional optical storage. [ABSTRACT FROM AUTHOR]

During the ignition process of a solid rocket motor, the pressure changes dramatically and the ignition process is very complex as it includes multiple reactions. Successful completion of the ignition process is essential for the proper operation of solid rocket motors. However, the measurement of pressure becomes extremely challenging due to several issues such as the enormity and high cost of conducting tests on solid rocket motors. Therefore, it needs to be investigated using numerical calculations and other methods. Currently, the fundamental theories concerning the ignition process have not been fully developed. In addition, numerical simulations require significant simplifications. To address these issues, this study proposes a solid rocket motor pressure prediction method based on bidirectional long short-term memory (CBiLSTM) combined with adaptive Gaussian noise (AGN). The method utilizes experimental pressure data and simulated pressure data as inputs for co-training to predict pressure data under new operating conditions. By comparison, the AGN-CBiLSTM method has a higher prediction accuracy with a percentage error of 3.27% between the predicted and actual data. This method provides an effective way to evaluate the performance of solid rocket motors and has a wide range of applications in the aerospace field. [ABSTRACT FROM AUTHOR]

Urban emergency shelters are a special type of public service facility, and their planning and construction are directly related to the safety of urban residents' lives and property, as well as to the sustainable development of the city. Existing research on the site selection of shelters is not ideal when dealing with large‐scale scenarios and fails to accurately reflect the actual situation. To address this issue, this study proposes an emergency shelter site selection model based on deep reinforcement learning, IAM‐PPO. This model constructs the site selection problem as a Markov Decision Process and uses deep learning to extract information from the site selection scenario. It finds the final solution for shelter locations through continuous exploration and learning. To improve the training efficiency of the model, the action masking process is innovatively applied to the model. The research results and ablation experiments using Shanghai as a case study prove that, owing to the diversity of shelter service ranges and action masking mechanism, the model proposed in this study can provide efficient and accurate shelter location services. Moreover, the customizability of this model provides meaningful reference value for other public facility location problems. [ABSTRACT FROM AUTHOR]

Background: Accurate pancreas and pancreatic tumor segmentation from abdominal scans is crucial for diagnosing and treating pancreatic diseases. Automated and reliable segmentation algorithms are highly desirable in both clinical practice and research. Purpose: Segmenting the pancreas and tumors is challenging due to their low contrast, irregular morphologies, and variable anatomical locations. Additionally, the substantial difference in size between the pancreas and small tumors makes this task difficult. This paper proposes an attention‐enhanced multiscale feature fusion network (AMFF‐Net) to address these issues via 3D attention and multiscale context fusion methods. Methods: First, to prevent missed segmentation of tumors, we design the residual depthwise attention modules (RDAMs) to extract global features by expanding receptive fields of shallow layers in the encoder. Second, hybrid transformer modules (HTMs) are proposed to model deep semantic features and suppress irrelevant regions while highlighting critical anatomical characteristics. Additionally, the multiscale feature fusion module (MFFM) fuses adjacent top and bottom scale semantic features to address the size imbalance issue. Results: The proposed AMFF‐Net was evaluated on the public MSD dataset, achieving 82.12% DSC for pancreas and 57.00% for tumors. It also demonstrated effective segmentation performance on the NIH and private datasets, outperforming previous State‐Of‐The‐Art (SOTA) methods. Ablation studies verify the effectiveness of RDAMs, HTMs, and MFFM. Conclusions: We propose an effective deep learning network for pancreas and tumor segmentation from abdominal CT scans. The proposed modules can better leverage global dependencies and semantic information and achieve significantly higher accuracy than the previous SOTA methods. [ABSTRACT FROM AUTHOR]

Individual Tree Detection-and-Counting (ITDC) is among the important tasks in town areas, and numerous methods are proposed in this direction. Despite their many advantages, still, the proposed methods are inadequate to provide robust results because they mostly rely on the direct field investigations. This paper presents a novel approach involving high-resolution imagery and the Canopy-Height-Model (CHM) data to solve the ITDC problem. The new approach is studied in six urban scenes: farmland, woodland, park, industrial land, road and residential areas. First, it identifies tree canopy regions using a deep learning network from high-resolution imagery. It then deploys the CHM-data to detect treetops of the canopy regions using a local maximum algorithm and individual tree canopies using the region growing. Finally, it calculates and describes the number of individual trees and tree canopies. The proposed approach is experimented with the data from Shanghai, China. Our results show that the individual tree detection method had an average overall accuracy of 0.953, with a precision of 0.987 for woodland scene. Meanwhile, the R2 value for canopy segmentation in different urban scenes is greater than 0.780 and 0.779 for canopy area and diameter size, respectively. These results confirm that the proposed method is robust enough for urban tree planning and management. [ABSTRACT FROM AUTHOR]

Introduction: Immune dysregulation plays a major role in cancer progression. The quantification of lymphocytic spatial inflammation may enable spatial system biology, improve understanding of therapeutic resistance, and contribute to prognostic imaging biomarkers. Methods: In this paper, we propose a knowledge-guided deep learning framework to measure the lymphocytic spatial architecture on human H&E tissue, where the fidelity of training labels is maximized through single-cell resolution image registration of H&E to IHC. We demonstrate that such an approach enables pixel-perfect ground-truth labeling of lymphocytes on H&E as measured by IHC. We then experimentally validate our technique in a genetically engineered, immune-compromised Rag2 mouse model, where Rag2 knockout mice lacking mature lymphocytes are used as a negative experimental control. Such experimental validation moves beyond the classical statistical testing of deep learning models and demonstrates feasibility of more rigorous validation strategies that integrate computational science and basic science. Results: Using our developed approach, we automatically annotated more than 111,000 human nuclei (45,611 CD3/CD20 positive lymphocytes) on H&E images to develop our model, which achieved an AUC of 0.78 and 0.71 on internal hold-out testing data and external testing on an independent dataset, respectively. As a measure of the global spatial architecture of the lymphocytic microenvironment, the average structural similarity between predicted lymphocytic density maps and ground truth lymphocytic density maps was 0.86 ± 0.06 on testing data. On experimental mouse model validation, we measured a lymphocytic density of 96.5 ± %1% in a Rag2+/- control mouse, compared to an average of 16.2 ± %5% in Rag2-/- immune knockout mice (p<0.0001, ANOVA-test). Discussion: These results demonstrate that CD3/CD20 positive lymphocytes can be accurately detected and characterized on H&E by deep learning and generalized across species. Collectively, these data suggest that our understanding of complex biological systems may benefit from computationally-derived spatial analysis, as well as integration of computational science and basic science. [ABSTRACT FROM AUTHOR]

Product configurators are recognised as critical toolkits enabling customers to co-create products with companies. Most available product configurators require customers to select suitable product attributes from predefined options. However, customers usually find the selection processes frustrating due to their lack of product knowledge. In view of the fact that customers often express their needs in imprecise and vague natural language, we define a new needs-based configuration mechanism and propose an implementation approach based on text embeddings and multilayer perceptron. Specifically, we leverage the massive amount of product reviews by encoding them into text embeddings. A multilayer perceptron is trained to map text embeddings to product attribute options. Experiment results indicate that the mapping has good generalisation capability to map customer needs into product configurations. The performance of our approach is comparable to that of deep learning-based approaches but with much higher efficiency in terms of computational complexity. Our needs-based configuration thus provides a quick and effective means of facilitating product customisation. It also demonstrates an innovative way of utilising customer resources in unstructured text to co-create products with companies. [ABSTRACT FROM AUTHOR]

In recent years, deep learning technology has developed significantly in the judicial field. More and more scholars are introducing deep learning to solve problems in the judicial field and improve the efficiency of judicial agencies in handling cases. An important task of judicial work is to identify the key elements of complex cases, establish the relationship between entities in the document, accurately grasp the development of the case, provide a basis for the understanding, analysis, and ruling of the case, and increase the interpretability of the case results. However, judging from the actual application situation, the application of deep learning technology in the judicial field is not yet mature and still faces huge challenges. On the one hand, the existing legal text element identification methods only consider the identification of core elements and ignore the basic elements. However, the basic elements contain necessary legal-related information, which is basic case information unique to the legal field and has important reference value for judicial staff to analyze cases. On the other hand, the traditional method of identifying key elements does not consider the contextual semantic relationship and ignores the element semantic information and global semantic information is lost, resulting in poor recognition of key elements of legal text. Therefore, in the identification of key elements of legal texts, not only the core elements and basic elements must be considered, but also the semantic features of sentences and the semantic features of elements must be comprehensively considered, thereby improving the effect of identifying key elements of legal texts and promoting the efficiency of judicial organs in handling cases. Aiming at the problem of identifying the basic elements of legal texts, this paper proposes a model for identifying the basic elements of legal texts based on dynamic representation. This algorithm uses the dynamic representation capability of the BERT model to vectorize text and considers global domain semantic information during the pre-training process to achieve a more comprehensive vector representation. Secondly, the memory component of the bidirectional long short-term memory network is used to integrate all the feature information between long-distance words in the legal text, effectively express the meaning of the word in the context, and realize the recognition of the basic elements of the legal text. In addition, the conditional random field model is used to learn the transfer rules of labels between entities and output the label sequence that best conforms to the actual rules. Finally, the basic element identification model proposed in this article has been significantly improved by comparing it with the current better methods. [ABSTRACT FROM AUTHOR]

Mass spectrometry is crucial in proteomics analysis, particularly using Data Independent Acquisition (DIA) for reliable and reproducible mass spectrometry data acquisition, enabling broad mass-to-charge ratio coverage and high throughput. DIA-NN, a prominent deep learning software in DIA proteome analysis, generates peptide results but may include low-confidence peptides. Conventionally, biologists have to manually screen peptide fragment ion chromatogram peaks (XIC) for identifying high-confidence peptides, a time-consuming and subjective process prone to variability. In this study, we introduce SeFilter-DIA, a deep learning algorithm, aiming at automating the identification of high-confidence peptides. Leveraging compressed excitation neural network and residual network models, SeFilter-DIA extracts XIC features and effectively discerns between high and low-confidence peptides. Evaluation of the benchmark datasets demonstrates SeFilter-DIA achieving 99.6% AUC on the test set and 97% for other performance indicators. Furthermore, SeFilter-DIA is applicable for screening peptides with phosphorylation modifications. These results demonstrate the potential of SeFilter-DIA to replace manual screening, providing an efficient and objective approach for high-confidence peptide identification while mitigating associated limitations. [ABSTRACT FROM AUTHOR]

The article introduces a new model called APF-GAN that improves upon existing generative adversarial network (GAN)-based models for image semantic segmentation. The model incorporates various strategies such as asymmetric pre-training and fine-tuning, progressive growth training, and semantic spatially-guided noise to enhance image quality and training efficiency. Experimental results demonstrate that the APF-GAN outperforms other models in terms of segmentation accuracy and aesthetics. The article also provides examples of images generated by the APF-GAN and discusses potential areas for future research. [Extracted from the article]

Machine learning (ML) has been broadly applied for vadose zone applications in recent years. This article provides a comprehensive review of such developments. ML applications for variables corresponding to different complex vadose zone processes are summarized mostly in a prediction context. By analyzing and assessing these applications, we discovered extensive usages of classic ML models with relatively limited applications of deep learning (DL) approaches in general. We also recognized a lack of benchmark datasets for soil property research as well as limited integration of physics‐based vadose zone principles into the ML approaches. To facilitate this interdisciplinary research of ML in vadose zone characterization and processes, a paradigm of knowledge‐guided machine learning is suggested along with other data‐driven and ML model‐based research suggestions to advance future research. Core Ideas: Random forest and artificial neural network are two widely applied machine learning options for predicting vadose zone studies.A benchmark dataset is missing in soil property studies.We suggest vadose zone scientists explore more deep learning options and expand knowledge‐guided machine learning implementations. [ABSTRACT FROM AUTHOR]

The challenge of spatial resource allocation is pervasive across various domains such as transportation, industry, and daily life. As the scale of real-world issues continues to expand and demands for real-time solutions increase, traditional algorithms face significant computational pressures, struggling to achieve optimal efficiency and real-time capabilities. In recent years, with the escalating computational power of computers, the remarkable achievements of reinforcement learning in domains like Go and robotics have demonstrated its robust learning and sequential decision-making capabilities. Given these advancements, there has been a surge in novel methods employing reinforcement learning to tackle spatial resource allocation problems. These methods exhibit advantages such as rapid solution convergence and strong model generalization abilities, offering a new perspective on resolving spatial resource allocation problems. Despite the progress, reinforcement learning still faces hurdles when it comes to spatial resource allocation. There remains a gap in its ability to fully grasp the diversity and intricacy of real-world resources. The environmental models used in reinforcement learning may not always capture the spatial dynamics accurately. Moreover, in situations laden with strict and numerous constraints, reinforcement learning can sometimes fall short in offering feasible strategies. Consequently, this paper is dedicated to summarizing and reviewing current theoretical approaches and practical research that utilize reinforcement learning to address issues pertaining to spatial resource allocation. In addition, the paper accentuates several unresolved challenges that urgently necessitate future focus and exploration within this realm and proposes viable approaches for these challenges. This research furnishes valuable insights that may assist scholars in gaining a more nuanced understanding of the problems, opportunities, and potential directions concerning the application of reinforcement learning in spatial resource allocation. [ABSTRACT FROM AUTHOR]

Background: In radiotherapy, real‐time tumor tracking can verify tumor position during beam delivery, guide the radiation beam to target the tumor, and reduce the chance of a geometric miss. Markerless kV x‐ray image‐based tumor tracking is challenging due to the low tumor visibility caused by tumor‐obscuring structures. Developing a new method to enhance tumor visibility for real‐time tumor tracking is essential. Purpose: To introduce a novel method for markerless kV image‐based tracking of lung tumors via deep learning‐based target decomposition. Methods: We utilized a conditional Generative Adversarial Network (cGAN), known as Pix2Pix, to build a patient‐specific model and generate the synthetic decomposed target image (sDTI) to enhance tumor visibility on the real‐time kV projection images acquired by the onboard kV imager equipped on modern linear accelerators. We used 4DCT simulation images to generate the digitally reconstructed radiograph (DRR) and DTI image pairs for model training. We augmented the training dataset by randomly shifting the 4DCT in the superior‐inferior, anterior‐posterior, and left‐right directions during the DRR and DTI generation process. We performed real‐time 2D tumor tracking via template matching between the DTI generated from the CT simulation and the sDTI generated from the real‐time kV projection images. We validated the proposed method using nine patients' datasets with implanted beacons near the tumor. Results: The sDTI can effectively improve the image contrast around the lung tumors on the kV projection images for the nine patients. With the beacon motion as ground truth, the tracking errors were on average 0.8 ± 0.7 mm in the superior‐inferior (SI) direction and 0.9 ± 0.8 mm in the in‐plane left‐right (IPLR) direction. The percentage of successful tracking, defined as a tracking error less than 2 mm in the SI direction, is 92.2% on the 4312 tested images. The patient‐specific model took approximately 12 h to train. During testing, it took approximately 35 ms to generate one sDTI, and 13 ms to perform the tumor tracking using template matching. Conclusions: Our method offers the potential solution for nearly real‐time markerless lung tumor tracking. It achieved a high level of accuracy and an impressive tracking rate. Further development of 3D lung tumor tracking is warranted. [ABSTRACT FROM AUTHOR]

Recent machine learning models for bandgap prediction that explicitly encode the structure information to the model feature set significantly improve the model accuracy compared to both traditional machine learning and non-graph-based deep learning methods. The ongoing rapid growth of open-access bandgap databases can benefit such model construction not only by expanding their domain of applicability but also by requiring constant updating of the model. Here, we build a new state-of-the-art multi-fidelity graph network model for bandgap prediction of crystalline compounds from a large bandgap database of experimental and density functional theory (DFT) computed bandgaps with over 806 600 entries (1500 experimental, 775 700 low-fidelity DFT, and 29 400 high-fidelity DFT). The model predicts bandgaps with a 0.23 eV mean absolute error in cross validation for high-fidelity data, and including the mixed data from all different fidelities improves the prediction of the high-fidelity data. The prediction error is smaller for high-symmetry crystals than for low symmetry crystals. Our data are published through a new cloud-based computing environment, called the "Foundry," which supports easy creation and revision of standardized data structures and will enable cloud accessible containerized models, allowing for continuous model development and data accumulation in the future. [ABSTRACT FROM AUTHOR]

Various studies have been conducted on applying intelligent recognition technology, especially speech recognition technology to improve English learning ability, mostly listening and speaking. However, few studies have touched on how image-to-text recognition technology can be used for writing. The present research was conducted to fill this gap by exploring the optimization of a deep-learning-based image captioning algorithm to facilitate English writing, so as to enable learners to break the time and space limitations and learn English writing (including sentence patterns, spelling, vocabulary, and grammar) anytime and anywhere by taking pictures. Therefore, this paper focused on image captioning based on CNN(Convolutional Neural Networks) and LSTM(Long Short-Term Memory), using DenseNet201 or Vision Transformer trained on the ImageNet-1K image classification dataset as the image encoder and LSTM as the decoder. First, pre-training was performed on the Flickr8k dataset. After selecting the best-trained model as the pre-trained weight model for the COCO dataset, fine-tuning optimization was performed on the COCO dataset, and the attention mechanism was used to design the ablation experiment. The BLEU-4, CIDEr, METEOR, and ROUGE evaluation indexes of the optimized model on the test set were 0.3437,1.121,0.2750, and 0.5117, respectively. The study results showed that the convergence of the model was accelerated and had better performance. The model was used to automatically caption 12 images that had never been used during the training process. The descriptions generated by the optimized image captioning algorithm have lexical and syntactic accuracy, and matched what the images expressed, showing that this improved algorithm could be used as a learning tool to help English learners improve lexical and syntactic acquisition to promote writing through the generated descriptions of the pictures taken anytime and anywhere in real-life situations. [ABSTRACT FROM AUTHOR]

Objectives: To evaluate the performance of automatic deep learning (DL) algorithm for size, mass, and volume measurements in predicting prognosis of lung adenocarcinoma (LUAD) and compared with manual measurements. Methods: A total of 542 patients with clinical stage 0-I peripheral LUAD and with preoperative CT data of 1-mm slice thickness were included. Maximal solid size on axial image (MSSA) was evaluated by two chest radiologists. MSSA, volume of solid component (SV), and mass of solid component (SM) were evaluated by DL. Consolidation-to-tumor ratios (CTRs) were calculated. For ground glass nodules (GGNs), solid parts were extracted with different density level thresholds. The prognosis prediction efficacy of DL was compared with that of manual measurements. Multivariate Cox proportional hazards model was used to find independent risk factors. Results: The prognosis prediction efficacy of T-staging (TS) measured by radiologists was inferior to that of DL. For GGNs, MSSA-based CTR measured by radiologists (RMSSA%) could not stratify RFS and OS risk, whereas measured by DL using 0HU (2D−AIMSSA0HU%) could by using different cutoffs. SM and SV measured by DL using 0 HU (AISM0HU% and AISV0HU%) could effectively stratify the survival risk regardless of different cutoffs and were superior to 2D−AIMSSA0HU%. AISM0HU% and AISV0HU% were independent risk factors. Conclusion: DL algorithm can replace human for more accurate T-staging of LUAD. For GGNs, 2D−AIMSSA0HU% could predict prognosis rather than RMSSA%. The prediction efficacy of AISM0HU% and AISV0HU% was more accurate than of 2D−AIMSSA0HU% and both were independent risk factors. Clinical relevance statement: Deep learning algorithm could replace human for size measurements and could better stratify prognosis than manual measurements in patients with lung adenocarcinoma. Key Points: • Deep learning (DL) algorithm could replace human for size measurements and could better stratify prognosis than manual measurements in patients with lung adenocarcinoma (LUAD). • For GGNs, maximal solid size on axial image (MSSA)–based consolidation-to-tumor ratio (CTR) measured by DL using 0 HU could stratify survival risk than that measured by radiologists. • The prediction efficacy of mass- and volume-based CTRs measured by DL using 0 HU was more accurate than of MSSA-based CTR and both were independent risk factors. [ABSTRACT FROM AUTHOR]

Realistic image composition aims to composite new images by fusing a source object into a target image. It is a challenging problem due to the complex multi-task framework, including sensible object placement, appearance consistency, shadow generation, etc. Most existing researchers attempt to address one of the issues. Especially before compositing, there is no matching assignment between the source object and target image, which often leads to unreasonable results. To address the issues above, we consider image composition as an image generation problem and propose a deep adversarial learning network via spatial position analysis. We target the analysis network segment and classify the objects in target images. A spatial alignment network matches the segmented objects with the source objects, and predicts a sensible placement position, and an adversarial network generates a realistic composite image with the shadow and reflection of the source object. Furthermore, we use the classification information of target objects to filter out unreasonable image compositing. Moreover, we introduce a new test set to evaluate the network generalization for our multi-task image composition dataset. Extensive experimental results of the SHU (Shanghai University) dataset demonstrate that our deep spatial position analysis network remarkably enhances the compositing performance in realistic, shadow, and reflection generations. [ABSTRACT FROM AUTHOR]

SUMMARY: Spartina alterniflora is a halophyte that can survive in high‐salinity environments, and it is phylogenetically close to important cereal crops, such as maize and rice. It is of scientific interest to understand why S. alterniflora can live under such extremely stressful conditions. The molecular mechanism underlying its high‐saline tolerance is still largely unknown. Here we investigated the possibility that high‐affinity K+ transporters (HKTs), which function in salt tolerance and maintenance of ion homeostasis in plants, are responsible for salt tolerance in S. alterniflora. To overcome the imprecision and unstable of the gene screening method caused by the conventional sequence alignment, we used a deep learning method, DeepGOPlus, to automatically extract sequence and protein characteristics from our newly assemble S. alterniflora genome to identify SaHKTs. Results showed that a total of 16 HKT genes were identified. The number of S. alterniflora HKTs (SaHKTs) is larger than that in all other investigated plant species except wheat. Phylogenetically related SaHKT members had similar gene structures, conserved protein domains and cis‐elements. Expression profiling showed that most SaHKT genes are expressed in specific tissues and are differentially expressed under salt stress. Yeast complementation expression analysis showed that type I members SaHKT1;2, SaHKT1;3 and SaHKT1;8 and type II members SaHKT2;1, SaHKT2;3 and SaHKT2;4 had low‐affinity K+ uptake ability and that type II members showed stronger K+ affinity than rice and Arabidopsis HKTs, as well as most SaHKTs showed preference for Na+ transport. We believe the deep learning‐based methods are powerful approaches to uncovering new functional genes, and the SaHKT genes identified are important resources for breeding new varieties of salt‐tolerant crops. Significance Statement: The great potential of deep learning‐based methods for uncovering the HKT genes in S. alterniflora was demonstrated, which opens the door for utilizing SaHKTs in a wide range of species to increase the stress resilience of staple crops. [ABSTRACT FROM AUTHOR]

Ubiquitin-specific-processing protease 7 (USP7) is a promising target protein for cancer therapy, and great attention has been given to the identification of USP7 inhibitors. Traditional virtual screening methods have now been successfully applied to discover USP7 inhibitors aiming at reducing costs and speeding up time in several studies. However, due to their unsatisfactory accuracy, it is still a difficult task to develop USP7 inhibitors. In this study, multiple supervised learning classifiers were built to distinguish active USP7 inhibitors from inactive ligands. Physicochemical descriptors, MACCS keys, ECFP4 fingerprints and SMILES were first calculated to represent the compounds in our in-house dataset. Two deep learning (DL) models and nine classical machine learning (ML) models were then constructed based on different combinations of the above molecular representations under three activity cutoff values, and a total of 15 groups of experiments (75 experiments) were implemented. The performance of the models in these experiments was evaluated, compared and discussed using a variety of metrics. The optimal models are ensemble learning models when the dataset is balanced or severely imbalanced, and SMILES-based DL performs the best when the dataset is slightly imbalanced. Meanwhile, multimodal data fusion in some cases can improve the performance of ML and DL models. In addition, SMOTE, unbiased decoy selection and SMILES enumeration can improve the performance of ML and DL models when the dataset is severely imbalanced, and SMOTE works the best. Our study established highly accurate supervised learning classification models, which would accelerate the development of USP7 inhibitors. Some guidance was also provided for drug researchers in selecting supervised models and molecular representations as well as handling imbalanced datasets. [ABSTRACT FROM AUTHOR]

Until recently, conventional biochemical staining had the undisputed status as well-established benchmark for most biomedical problems related to clinical diagnostics, fundamental research and biotechnology. Despite this role as gold-standard, staining protocols face several challenges, such as a need for extensive, manual processing of samples, substantial time delays, altered tissue homeostasis, limited choice of contrast agents, 2D imaging instead of 3D tomography and many more. Label-free optical technologies, on the other hand, do not rely on exogenous and artificial markers, by exploiting intrinsic optical contrast mechanisms, where the specificity is typically less obvious to the human observer. Over the past few years, digital staining has emerged as a promising concept to use modern deep learning for the translation from optical contrast to established biochemical contrast of actual stainings. In this review article, we provide an in-depth analysis of the current state-of-the-art in this field, suggest methods of good practice, identify pitfalls and challenges and postulate promising advances towards potential future implementations and applications. [ABSTRACT FROM AUTHOR]

(1) Background: Cell proliferation (Ki-67) has important clinical value in the treatment and prognosis of non-small cell lung cancer (NSCLC). However, current detection methods for Ki-67 are invasive and can lead to incorrect results. This study aimed to explore a deep learning classification model for the prediction of Ki-67 and the prognosis of NSCLC based on FDG-PET/CT images. (2) Methods: The FDG-PET/CT scan results of 159 patients with NSCLC confirmed via pathology were analyzed retrospectively, and the prediction models for the Ki-67 expression level based on PET images, CT images and PET/CT combined images were constructed using Densenet201. Based on a Ki-67 high expression score (HES) obtained from the prediction model, the survival rate of patients with NSCLC was analyzed using Kaplan–Meier and univariate Cox regression. (3) Results: The statistical analysis showed that Ki-67 expression was significantly correlated with clinical features of NSCLC, including age, gender, differentiation state and histopathological type. After a comparison of the three models (i.e., the PET model, the CT model, and the FDG-PET/CT combined model), the combined model was found to have the greatest advantage in Ki-67 prediction in terms of AUC (0.891), accuracy (0.822), precision (0.776) and specificity (0.902). Meanwhile, our results indicated that HES was a risk factor for prognosis and could be used for the survival prediction of NSCLC patients. (4) Conclusions: The deep-learning-based FDG-PET/CT radiomics classifier provided a novel non-invasive strategy with which to evaluate the malignancy and prognosis of NSCLC. [ABSTRACT FROM AUTHOR]

Military, agricultural, and urban planning have all made extensive use of SAR/ISAR in the realm of remote sensing. SAR/ISAR images are more capable of identifying the details of the targets than optical images and can be taken in any condition. Due to the challenges associated with SAR/ISAR imaging, the lack of data causes many jobs relying on data-driven deep learning algorithms to perform less than satisfactorily. Cropping, rotation, and other procedures are examples of classic data augmentation techniques now in use, although they do not fundamentally differ from basic replication and cannot increase the model's stability and robustness. Deep generative models are used to generate SAR/ISAR images, which is a more efficient way than the conventional ones. The generation techniques are outlined and organized depending on the application fields in this review, including SAR/ISAR data augmentation (26 papers), SAR/ISAR image translation (29 papers), SAR/ISAR image enhancement (22 papers), azimuth interpolation (9 papers), and deceptive jamming (1 paper). The connected works are then summarized based on several deep generative models. 87 linked studies and 5 associated survey papers from 2017 to 2022 are compiled in this review. Finally, the summarized works are systematically analyzed. There are 27 papers using MSTAR for image generation, which is the mostly applied dataset. For evaluation, the combination of SSIM and PSNR is applied most widely (32.19%). In conclusion, this review offers fresh perspectives on the direction in which deep generative models for SAR/ISAR image generation are headed. The cutting-edge methods outlined in this paper are also available to researchers in other domains. [ABSTRACT FROM AUTHOR]

Land subsidence is a geohazard phenomenon caused by the lowering of land elevation due to the compression of the sinking land soil body, thus creating an excessive constraint on the safe construction and sustainable development of cities. The use of accurate and efficient means for land subsidence prediction is of remarkable importance for preventing land subsidence and ensuring urban safety. Although the current time-series prediction method can accomplish relatively high accuracy, the predicted settlement points are independent of each other, and the existence of spatial dependence in the data itself is lost. In order to unlock this problem, a spatial convolutional long short-term memory neural network (ConvLSTM) based on the spatio-temporal prediction method for land subsidence is constructed. To this end, a cloud platform is employed to obtain a long time series deformation dataset from May 2017 to November 2021 in the understudied area. A convolutional structure to extract spatial features is utilized in the proposed model, and an LSTM structure is linked to the model for time-series prediction to achieve unified modeling of temporal and spatial correlation, thereby rationally predicting the land subsidence progress trend and distribution. The experimental results reveal that the prediction results of the ConvLSTM model are more accurate than those of the LSTM in about 62% of the understudied area, and the overall mean absolute error (MAE) is reduced by about 7%. The achieved results exhibit better prediction in the subsidence center region, and the spatial distribution characteristics of the subsidence data are effectively captured. The present prediction results are more consistent with the distribution of real subsidence and could provide more accurate and reasonable scientific references for subsidence prevention and control in the Beijing-Tianjin-Hebei region. [ABSTRACT FROM AUTHOR]

This paper proposes a novel domain-generation-algorithm detection framework based on statistical learning that integrates the detection capabilities of multiple heterogeneous models. The framework includes both traditional machine learning methods based on artificial features and deep learning methods, comprehensively analyzing 34 artificial features and advanced features extracted from deep neural networks. Additionally, the framework evaluates the predictions of the base models based on the fit of the samples to each type of sample set and a predefined significance level. The predictions of the base models are statistically analyzed, and the final decision is made using strategies such as voting, confidence, and credibility. Experimental results demonstrate that the DGA detection framework based on statistical learning achieves a higher detection rate compared to the underlying base models, with accuracy, precision, recall, and F1 scores reaching 0.979, 0.977, 0.981, and 0.979, respectively. The framework also exhibits a stronger adaptability to unknown domains and a certain level of robustness against concept drift attacks. [ABSTRACT FROM AUTHOR]

In the era of deep learning, representational text-matching algorithms based on BERT and its variant models have become mainstream and are limited by the sentence vectors generated by the BERT model, and the SimCSE algorithm proposed in 2021 has improved the sentence vector quality to a certain extent. In this paper, to address the problem that the SimCSE algorithm has—that the greater the difference in sentence length, the smaller the probability that the sentence pairs are similar—an EdaCSE algorithm is proposed to perturb the sentence length using a simple data enhancement method without affecting the semantics of the sentences. The perturbation is applied to the sentence length by adding meaningless English punctuation marks to the original sentence so that the model no longer tends to recognise sentences of similar length as similar sentences. Based on the BERT series of models, experiments were conducted on five different datasets, and the experiments proved that the EdaCSE method improves an average of 1.67, 0.84, and 1.08 on the five datasets. [ABSTRACT FROM AUTHOR]

Background: X‐ray image quality is critical for accurate intrafraction motion tracking in radiation therapy. Purpose: This study aims to develop a deep‐learning algorithm to improve kV image contrast by decomposing the image into bony and soft tissue components. In particular, we designed a priori attention mechanism in the neural network framework for optimal decomposition. We show that a patient‐specific prior cross‐attention (PCAT) mechanism can boost the performance of kV image decomposition. We demonstrate its use in paraspinal SBRT motion tracking with online kV imaging. Methods: Online 2D kV projections were acquired during paraspinal SBRT for patient motion monitoring. The patient‐specific prior images were generated by randomly shifting and rotating spine‐only DRR created from the setup CBCT, simulating potential motions. The latent features of the prior images were incorporated into the PCAT using multi‐head cross attention. The neural network aimed to learn to selectively amplify the transmission of the projection image features that correlate with features of the priori. The PCAT network structure consisted of (1) a dual‐branch generator that separates the spine and soft tissue component of the kV projection image and (2) a dual‐function discriminator (DFD) that provides the realness score of the predicted images. For supervision, we used a loss combining mean absolute error loss, discriminator loss, perceptual loss, total variation, and mean squared error loss for soft tissues. The proposed PCAT approach was benchmarked against previous work using the ResNet generative adversarial network (ResNetGAN) without prior information. Results: The trained PCAT had improved performance in effectively retaining and preserving the spine structure and texture information while suppressing the soft tissues from the kV projection images. The decomposed spine‐only x‐ray images had the submillimeter matching accuracy at all beam angles. The decomposed spine‐only x‐ray significantly reduced the maximum errors to 0.44 mm (<2 pixels) in comparison to 0.92 mm (∼4 pixels) of ResNetGAN. The PCAT decomposed spine images also had higher PSNR and SSIM (p‐value < 0.001). Conclusion: The PCAT selectively learned the important latent features by incorporating the patient‐specific prior knowledge into the deep learning algorithm, significantly improving the robustness of the kV projection image decomposition, and leading to improved motion tracking accuracy in paraspinal SBRT. [ABSTRACT FROM AUTHOR]

In recent years, artificial intelligence (AI) technology has been used in most if not all domains and has greatly benefited our lives. While AI can accurately extract critical features and valuable information from large amounts of data to help people complete tasks faster, there are growing concerns about the non-transparency of AI in the decision-making process. The emergence of explainable AI (XAI) has allowed humans to better understand and control AI systems, which is motivated to provide transparent explanations for the decisions made by AI. This article aims to present a comprehensive overview of recent research on XAI approaches from three well-defined taxonomies. We offer an in-depth analysis and summary of the status and prospects of XAI applications in several key areas where reliable explanations are urgently needed to avoid mistakes in decision-making. We conclude by discussing XAI's limitations and future research directions. [ABSTRACT FROM AUTHOR]

Aiming at the difficulty in locating the picking points of table grapes in the natural environment, this paper proposed a Yolov5-CBAM-Fourth Detection Layer-Decoupled (Yolov5-CFD) network based on Yolov5 to realize grape and stem recognition. Meanwhile, the fast localization of the picking point was realized using a geometric method. First, to enhance the feature extraction ability of the Backbone module, it was improved by integrating the Convolutional Block Attention Module (CBAM) attention mechanism. Second, aiming at the problem that a small target was difficult to detect, the fourth layer of detection was added to the Neck module. In addition, the paper improved the Head module of Yolov5 by borrowing from Decoupled structure in Yolox, which optimized the classification and regression performance of the network. Further, a geometric method was used to quickly and accurately locate the picking points of table grapes. In order to verify the effectiveness of the proposed network model, about 10,000 grape images were used for training. The results showed that the detection precision, recall, mAP_0.5 and mAP_0.5:0.95 of the Yolov5s-CFD model were 0.857, 0.804, 0.855, 0.642, respectivaly. And the detection precision, recall, mAP_0.5 and mAP_0.5:0.95 of the Yolov5m-CFD model were 0.986, 0.987, 0.993 and 0.910, respectivaly. In addition, the success rate of picking point localization was compared with the corresponding network structure. The results showed that the success rate of picking point positioning of the Yolov5s-CFD model was increased by 11.53% compared with the initial Yolov5s, and the success rate of the Yolov5m-CFD model was increased by 5.84% compared with the initial Yolov5. Although the recognition time of the improved Yolov5 model is increased compared to the initial Yolov5 model, it is still acceptable. It can fully meet the requirements of mechanized picking of table grapes and provide a theoretical basis for mechanized picking of table grapes. [ABSTRACT FROM AUTHOR]

The optical microscope is customarily an instrument of substantial size and expense but limited performance. Here we report an integrated microscope that achieves optical performance beyond a commercial microscope with a 5×, NA 0.1 objective but only at 0.15 cm3 and 0.5 g, whose size is five orders of magnitude smaller than that of a conventional microscope. To achieve this, a progressive optimization pipeline is proposed which systematically optimizes both aspherical lenses and diffractive optical elements with over 30 times memory reduction compared to the end-to-end optimization. By designing a simulation-supervision deep neural network for spatially varying deconvolution during optical design, we accomplish over 10 times improvement in the depth-of-field compared to traditional microscopes with great generalization in a wide variety of samples. To show the unique advantages, the integrated microscope is equipped in a cell phone without any accessories for the application of portable diagnostics. We believe our method provides a new framework for the design of miniaturized high-performance imaging systems by integrating aspherical optics, computational optics, and deep learning. Traditional optical microscope, while bulky, often fails to deliver optimal performance. Here, the authors have engineered an integrated microscope of 0.15 cm3 in volume and a weight of 0.5 g, which outperforms a commercial microscope and can be seamlessly integrated with a smartphone. [ABSTRACT FROM AUTHOR]

Background: Due to intrinsic differences in data formatting, data structure, and underlying semantic information, the integration of imaging data with clinical data can be non‐trivial. Optimal integration requires robust data fusion, that is, the process of integrating multiple data sources to produce more useful information than captured by individual data sources. Here, we introduce the concept of fusion quality for deep learning problems involving imaging and clinical data. We first provide a general theoretical framework and numerical validation of our technique. To demonstrate real‐world applicability, we then apply our technique to optimize the fusion of CT imaging and hepatic blood markers to estimate portal venous hypertension, which is linked to prognosis in patients with cirrhosis of the liver. Purpose: To develop a measurement method of optimal data fusion quality deep learning problems utilizing both imaging data and clinical data. Methods: Our approach is based on modeling the fully connected layer (FCL) of a convolutional neural network (CNN) as a potential function, whose distribution takes the form of the classical Gibbs measure. The features of the FCL are then modeled as random variables governed by state functions, which are interpreted as the different data sources to be fused. The probability density of each source, relative to the probability density of the FCL, represents a quantitative measure of source‐bias. To minimize this source‐bias and optimize CNN performance, we implement a vector‐growing encoding scheme called positional encoding, where low‐dimensional clinical data are transcribed into a rich feature space that complements high‐dimensional imaging features. We first provide a numerical validation of our approach based on simulated Gaussian processes. We then applied our approach to patient data, where we optimized the fusion of CT images with blood markers to predict portal venous hypertension in patients with cirrhosis of the liver. This patient study was based on a modified ResNet‐152 model that incorporates both images and blood markers as input. These two data sources were processed in parallel, fused into a single FCL, and optimized based on our fusion quality framework. Results: Numerical validation of our approach confirmed that the probability density function of a fused feature space converges to a source‐specific probability density function when source data are improperly fused. Our numerical results demonstrate that this phenomenon can be quantified as a measure of fusion quality. On patient data, the fused model consisting of both imaging data and positionally encoded blood markers at the theoretically optimal fusion quality metric achieved an AUC of 0.74 and an accuracy of 0.71. This model was statistically better than the imaging‐only model (AUC = 0.60; accuracy = 0.62), the blood marker‐only model (AUC = 0.58; accuracy = 0.60), and a variety of purposely sub‐optimized fusion models (AUC = 0.61–0.70; accuracy = 0.58–0.69). Conclusions: We introduced the concept of data fusion quality for multi‐source deep learning problems involving both imaging and clinical data. We provided a theoretical framework, numerical validation, and real‐world application in abdominal radiology. Our data suggests that CT imaging and hepatic blood markers provide complementary diagnostic information when appropriately fused. [ABSTRACT FROM AUTHOR]

At present, research on intelligent wheelchairs mostly focuses on motion control, while research on attitude-based adjustment is relatively insufficient. The existing methods for adjusting wheelchair posture generally lack collaborative control and good human–machine collaboration. This article proposes an intelligent wheelchair posture-adjustment method based on action intention recognition by studying the relationship between the force changes on the contact surface between the human body and the wheelchair and the action intention. This method is applied to a multi-part adjustable electric wheelchair, which is equipped with multiple force sensors to collect pressure information from various parts of the passenger's body. The upper level of the system converts the pressure data into the form of a pressure distribution map, extracts the shape features using the VIT deep learning model, identifies and classifies them, and ultimately identifies the action intentions of the passengers. Based on different action intentions, the electric actuator is controlled to adjust the wheelchair posture. After testing, this method can effectively collect the body pressure data of passengers, with an accuracy of over 95% for the three common intentions of lying down, sitting up, and standing up. The wheelchair can adjust its posture based on the recognition results. By adjusting the wheelchair posture through this method, users do not need to wear additional equipment and are less affected by the external environment. The target function can be achieved with simple learning, which has good human–machine collaboration and can solve the problem of some people having difficulty adjusting the wheelchair posture independently during wheelchair use. [ABSTRACT FROM AUTHOR]

The reform of tertiary education teaching promotes teachers to adjust timely teaching plans based on students' learning feedback in order to improve teaching performance. Thefore, learning score prediction is a key issue in process of the reform of tertiary education teaching. With the development of information and management technologies, a lot of teaching data are generated as the scale of online and offline education expands. However, a teacher or educator does not have a comprehensive dataset in practice, which challenges his/her ability to predict the students' learning performance from the individual's viewpoint. How to overcome the drawbacks of small samples is an open issue. To this end, it is desirable that an effective artificial intelligent tool is designed to help teachers or educators predict students' scores well. We propose a boosting and decision-tree-regression-based score prediction (BDTR-SP) model, which relies on an ensemble learning structure with base learners of decision tree regression (DTR) to improve the prediction accuracy. Experiments on small samples are conducted to examine the important features that affect students' scores. The results show that the proposed model has advantages over its peer in terms of prediction correctness. Moreover, the predicted results are consistent with the actual facts implied in the original dataset. The proposed BDTR-SP method aids teachers and students to predict students' performance in the on-going courses in order to adjust the teaching and learning strategies, plans and practices in advance, enhancing the teaching and learning quality. Therefore, the integration of information technology and artificial intelligence into teaching and learning practices is able to push forward the reform of tertiary education teaching. [ABSTRACT FROM AUTHOR]

Applied computing and artificial intelligence methods have been attracting growing interest in recent years due to their effectiveness in solving technical problems. Hao et al. [[14]] present an unsupervised fault diagnosis methodology to leverage the generated MPCs of different working conditions to diagnose the actual unlabeled MPCs. The paper by Ainapure et al. [[17]] proposes a new cross-domain fault diagnosis method with enhanced robustness. The paper by Saeed et al. [[1]] proposes an approach to building an AutoML data-dependent CNN model (DeepPCANet) customized for DR screening automatically. [Extracted from the article]

Digital systems storing cadastral data in vector format are considered effective due to their ability of offering interactive services to citizens and other land-related systems. The adoption of such systems is ubiquitous, but when adopted, they create two non-compatible systems with paper-based cadastral systems whose information needs to be digitised. This study proposes a new approach that is fast and accurate for transforming paper-based cadastral data into digital systems. The proposed method involves deep-learning techniques of the LCNN and ResNet-50 for detecting cadastral parcels and their numbers, respectively, from the cadastral plans. It also contains four functions defined to speed up transformations and compilations of the cadastral plan's data in digital systems. The LCNN is trained and validated with 968 samples. The ResNet-50 is trained and validated with 106,000 samples. The Structural-Average-Precision ( sAP 10 ) achieved with the LCNN was 0.9057. The Precision, Recall and F1-Score achieved with the ResNet-50 were 0.9650, 0.9648 and 0.9649, respectively. These results confirmed that the new method is accurate enough for implementation, and we tested it with a huge set of data from Tanzania. Its performance from the experimented data shows that the proposed method could effectively transform paper-based cadastral data into digital systems. [ABSTRACT FROM AUTHOR]

In the medical field, hematoxylin and eosin (H&E)-stained histopathology images of cell nuclei analysis represent an important measure for cancer diagnosis. The most valuable aspect of the nuclei analysis is the segmentation of the different nuclei morphologies of different organs and subsequent diagnosis of the type and severity of the disease based on pathology. In recent years, deep learning techniques have been widely used in digital histopathology analysis. Automated nuclear segmentation technology enables the rapid and efficient segmentation of tens of thousands of complex and variable nuclei in histopathology images. However, a challenging problem during nuclei segmentation is the blocking of cell nuclei, overlapping, and background complexity of the tissue fraction. To address this challenge, we present MIU-net, an efficient deep learning network structure for the nuclei segmentation of histopathology images. Our proposed structure includes two blocks with modified inception module and attention module. The advantage of the modified inception module is to balance the computation and network performance of the deeper layers of the network, combined with the convolutional layer using different sizes of kernels to learn effective features in a fast and efficient manner to complete kernel segmentation. The attention module allows us to extract small and fine irregular boundary features from the images, which can better segment cancer cells that appear disorganized and fragmented. We test our methodology on public kumar datasets and achieve the highest AUC score of 0.92. The experimental results show that the proposed method achieves better performance than other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Autonomous underwater vehicles (AUVs)-assisted mobile data collection in underwater wireless sensor networks (UWSNs) has received significant attention because of their mobility and flexibility. To satisfy the increasing demand of diverse application requirements for underwater data collection, such as time-sensitive data freshness, emergency event security as well as energy efficiency, in this paper, we propose a novel multi-modal AUV-assisted data collection scheme which integrates both acoustic and optical technologies and takes advantage of their complementary strengths in terms of communication distance and data rate. In this scheme, we consider the age of information (AoI) of the data packet, node transmission energy as well as energy consumption of the AUV movement, and we make a trade-off between them to retrieve data in a timely and reliable manner. To optimize these, we leverage a deep reinforcement learning (DRL) approach to find the optimal motion trajectory of AUV by selecting the suitable communication options. In addition to that, we also design an optimal angle steering algorithm for AUV navigation under different communication scenarios to reduce energy consumption further. We conduct extensive simulations to verify the effectiveness of the proposed scheme, and the results show that the proposed scheme can significantly reduce the weighted sum of AoI as well as energy consumption. [ABSTRACT FROM AUTHOR]

Background: Structural magnetic resonance imaging (sMRI) can provide morphological information about the structure and function of the brain in the same scanning process. It has been widely used in the diagnosis of Alzheimer's disease (AD) and mild cognitive impairment (MCI). Purpose: To capture the anatomical changes in the brain caused by AD/MCI, deep learning‐based MRI image analysis methods have been proposed in recent years. However, it is observed that the performance of most existing methods is limited as they only construct a single type of deep network and ignore the significance of other clinical information. Methods: To make up for these defects, an ensemble framework that incorporates three types of dedicatedly‐designed convolutional neural networks (CNNs) and a multilayer perceptron (MLP) network is proposed, where three CNNs with entropy‐based multi‐instance learning pooling layers have more reliable feature selection abilities. The dedicatedly‐designed base classifiers can make use of the heterogeneous data, and empower the framework with enhanced diversity and robustness. In particular, to consider the interactions among the base classifiers, a novel multi‐head self‐attention voting scheme is designed. Moreover, considering the chance that MCI can be transformed to AD, the proposed framework is designed to diagnose AD and predict MCI conversion simultaneously, with the aid of the transfer learning technique. Results: For performance evaluation and comparison, extensive experiments are conducted on the public dataset of the Alzheimer's Disease Neuroimaging Initiative (ADNI). The results show that the proposed ensemble framework provides superior performance under most of the evaluation metrics. Especially, the proposed framework achieves state‐of‐the‐art diagnostic accuracy (98.61% for the AD diagnosis task, and 84.49% for the MCI conversion prediction task). Conclusions: These promising results demonstrate the proposed ensemble framework can accurately diagnose AD patients and predict the conversion of MCI patients, which has the potential of clinical practice for diagnosing AD and MCI. [ABSTRACT FROM AUTHOR]

Simple Summary: HER2 expression is important for target therapy in breast cancer patients, however, accurate evaluation of HER2 expression is challenging for pathologists owing to the ambiguities and subjectivities of manual scoring. We proposed a deep learning framework using a Whole Slide gray value map and convolutional neural network model to predict HER2 expression level on immunohistochemistry (IHC) assay and predict HER2 gene status on fluorescence in situ hybridization (FISH) assay. Our results indicated that the proposed model is feasible for predicting HER2 expression and gene amplification and achieved high consistency with the experienced pathologists' assessment. This unique HER2 scoring model did not rely on challenging manual intervention and proved to be a simple and robust tool for pathologists to improve the accuracy of HER2 interpretation and provided a clinical aid to target therapy in breast cancer patients. Accurate detection of HER2 expression through immunohistochemistry (IHC) is of great clinical significance in the treatment of breast cancer. However, manual interpretation of HER2 is challenging, due to the interobserver variability among pathologists. We sought to explore a deep learning method to predict HER2 expression level and gene status based on a Whole Slide Image (WSI) of the HER2 IHC section. When applied to 228 invasive breast carcinoma of no special type (IBC-NST) DAB-stained slides, our GrayMap+ convolutional neural network (CNN) model accurately classified HER2 IHC level with mean accuracy 0.952 ± 0.029 and predicted HER2 FISH status with mean accuracy 0.921 ± 0.029. Our result also demonstrated strong consistency in HER2 expression score between our system and experienced pathologists (intraclass correlation coefficient (ICC) = 0.903, Cohen's κ = 0.875). The discordant cases were found to be largely caused by high intra-tumor staining heterogeneity in the HER2 IHC group and low copy number in the HER2 FISH group. [ABSTRACT FROM AUTHOR]

As a major Internet of Things information carrier within the greenhouse, image quality and image file size need to be balanced. To reduce image file size while maintaining image quality, a sub‐regional compression (SRC) method is proposed, and its optimal compression quality of the background regions is selected using image quality assessments such as SSIM, BRISQUE, NIQE, and PIQE. In addition, a convolutional neural network‐based NIMA model to measure image quality is also being introduced. In the end, the SRC method is compared with JPEG compression default mode and JPEG compression with quality 30 in NIMA score, compression time, and compression efficiency, proving the effectiveness of the method. Practical applications: Certain vegetables and fruits, such as cucumbers, tomatoes, etc., in most cases are processed directly into the hands of consumers, so the loss is directly attributable to their growth, especially pests and diseases. Monitoring for the above situation is a great improvement, but the increase in monitoring accuracy brings video/key frame to becoming larger. Therefore, it is better to reduce the file size as well as to ensure clarity in regions of interest, viewing the files faster and detecting the losses by humans or machines as easily as possible. This paper investigates the overall process and innovatively introduces deep learning into image quality assessment to better simulate human perception of images, and the research results of this paper can be applied to different models of surveillance cameras to achieve the same results. [ABSTRACT FROM AUTHOR]

Short-term energy consumption prediction of buildings is crucial for developing model-based predictive control, fault detection, and diagnosis methods. This study takes a university library in Xi'an as the research object. First, a time-by-time energy consumption prediction model is established under the supervised learning approach, which uses a long short-term memory (LSTM) network and a Multi-Input Multi-Output (MIMO) strategy. The experimental results validate the model's validity, which is close enough to physical reality for engineering purposes. Second, the potential of the people flows factor in energy consumption prediction models is explored. The results show that people flow has great potential in predicting building energy consumption and can effectively improve the prediction model performance. Third, a diagnostic method, which can recognize abnormal energy consumption data is used to diagnose the unreasonable use of the building during each hour of operation. The method is based on differences between actual and predicted energy consumption data derived from a short-term energy consumption prediction model. Based on actual building operation data, this work is enlightening and can serve as a reference for building energy efficiency management and operation. [ABSTRACT FROM AUTHOR]

Squeeze film damper (SFD) is one of the key components in the sophisticated rotating machines in the high-end aerospace industry. Rolling bearings in the aeroengine spindle system are often used in combination with elastic supports and SFDs to reduce rotor vibrations. Accurate coefficient identification of oil films in SFDs is critical to ensure safe and reliable operations of the complex equipment. However, this task remains a challenging issue in the past decades due to the difficulties in real-time measurement and computation during operations. The highly nonlinear relationship between the measurement and oil film coefficient cannot be effectively captured using traditional methods. This article proposes a novel deep-learning-based information fusion methodology for oil film coefficient identification of SFDs. An end-to-end deep neural network model structure is established, which explores heterogeneous SFD data simultaneously with fusion architecture and directly outputs the identification results. The high-dimensional data can be automatically processed, and high identification accuracy can be achieved. Experiments on the SFD test rig are carried out for validation. The experimental results demonstrate that the proposed method can well identify the oil film coefficients in an intelligent manner and is robust against environmental noise and different parameter settings. The proposed methodology is, thus, promising for applications in real industries. [ABSTRACT FROM AUTHOR]

Multi-view clustering (MVC) has attracted more and more attention in the recent few years by making full use of complementary and consensus information between multiple views to cluster objects into different partitions. Although there have been two existing works for MVC survey, neither of them jointly takes the recent popular deep learning-based methods into consideration. Therefore, in this paper, we conduct a comprehensive survey of MVC from the perspective of representation learning. It covers a quantity of multi-view clustering methods including the deep learning-based models, providing a novel taxonomy of the MVC algorithms. Furthermore, the representation learning-based MVC methods can be mainly divided into two categories, i.e., shallow representation learning-based MVC and deep representation learning-based MVC, where the deep learning-based models are capable of handling more complex data structure as well as showing better expression. In the shallow category, according to the means of representation learning, we further split it into two groups, i.e., multi-view graph clustering and multi-view subspace clustering. To be more comprehensive, basic research materials of MVC are provided for readers, containing introductions of the commonly used multi-view datasets with the download link and the open source code library. In the end, some open problems are pointed out for further investigation and development. [ABSTRACT FROM AUTHOR]

Purpose: Computer‐aided automatic pancreas segmentation is essential for early diagnosis and treatment of pancreatic diseases. However, the annotation of pancreas images requires professional doctors and considerable expenditure. Due to imaging differences among various institution population, scanning devices, imaging protocols, and so on, significant degradation in the performance of model inference results is prone to occur when models trained with domain‐specific (usually institution‐specific) datasets are directly applied to new (other centers/institutions) domain data. In this paper, we propose a novel unsupervised domain adaptation method based on adversarial learning to address pancreas segmentation challenges with the lack of annotations and domain shift interference. Methods: A 3D semantic segmentation model with attention module and residual module is designed as the backbone pancreas segmentation model. In both segmentation model and domain adaptation discriminator network, a multiscale progressively weighted structure is introduced to acquire different field of views. Features of labeled data and unlabeled data are fed in pairs into the proposed multiscale discriminator to learn domain‐specific characteristics. Then the unlabeled data features with pseudodomain label are fed to the discriminator to acquire domain‐ambiguous information. With this adversarial learning strategy, the performance of the segmentation network is enhanced to segment unseen unlabeled data. Results: Experiments were conducted on two public annotated datasets as source datasets, respectively, and one private dataset as target dataset, where annotations were not used for the training process but only for evaluation. The 3D segmentation model achieves comparative performance with state‐of‐the‐art pancreas segmentation methods on source domain. After implementing our domain adaptation architecture, the average dice similarity coefficient (DSC) of the segmentation model trained on the NIH‐TCIA source dataset increases from 58.79% to 72.73% on the local hospital dataset, while the performance of the target domain segmentation model transferred from the medical segmentation decathlon (MSD) source dataset rises from 62.34% to 71.17%. Conclusions: Correlations of features across data domains are utilized to train the pancreas segmentation model on unlabeled data domain, improving the generalization of the model. Our results demonstrate that the proposed method enables the segmentation model to make meaningful segmentation for unseen data of the training set. In the future, the proposed method has the potential to apply segmentation model trained on public dataset to clinical unannotated CT images from local hospital, effectively assisting radiologists in clinical practice. [ABSTRACT FROM AUTHOR]

Streamflow prediction is a long‐standing hydrologic problem. Development of models for streamflow prediction often requires incorporation of catchment physical descriptors to characterize the associated complex hydrological processes. Across different scales of catchments, these physical descriptors also allow models to extrapolate hydrologic information from one catchment to others, a process referred to as "regionalization". Recently, in gauged basin scenarios, deep learning models have been shown to achieve state of the art regionalization performance by building a global hydrologic model. These models predict streamflow given catchment physical descriptors and weather forcing data. However, these physical descriptors are by their nature uncertain, sometimes incomplete, or even unavailable in certain cases, which limits the applicability of this approach. In this paper, we show that by assigning a vector of random values as a surrogate for catchment physical descriptors, we can achieve robust regionalization performance under a gauged prediction scenario. Our results show that the deep learning model using our proposed random vector approach achieves a predictive performance comparable to that of the model using actual physical descriptors. The random vector approach yields robust performance under different data sparsity scenarios and deep learning model selections. Furthermore, based on the use of random vectors, high‐dimensional characterization improves regionalization performance in gauged basin scenario when physical descriptors are uncertain, or insufficient. Key Points: We propose a random vector strategy to serve as a surrogate to physical descriptors for regionalization under gauged scenariosThe deep learning model using random vectors yields comparable performance to the model using physical descriptors in gauged scenariosHigh‐dimensional characterization helps distinguish catchments and thus helps catchments learn from each other [ABSTRACT FROM AUTHOR]

Purpose: Stent has often been used as an internal surrogate to monitor intrafraction tumor motion during pancreatic cancer radiotherapy. Based on the stent contours generated from planning CT images, the current intrafraction motion review (IMR) system on Varian TrueBeam only provides a tool to verify the stent motion visually but lacks quantitative information. The purpose of this study is to develop an automatic stent recognition method for quantitative intrafraction tumor motion monitoring in pancreatic cancer treatment. Methods: A total of 535 IMR images from 14 pancreatic cancer patients were retrospectively selected in this study, with the manual contour of the stent on each image serving as the ground truth. We developed a deep learning–based approach that integrates two mechanisms that focus on the features of the segmentation target. The objective attention modeling was integrated into the U‐net framework to deal with the optimization difficulties when training a deep network with 2D IMR images and limited training data. A perceptual loss was combined with the binary cross‐entropy loss and a Dice loss for supervision. The deep neural network was trained to capture more contextual information to predict binary stent masks. A random‐split test was performed, with images of ten patients (71%, 380 images) randomly selected for training, whereas the rest of four patients (29%, 155 images) were used for testing. Sevenfold cross‐validation of the proposed PAUnet on the 14 patients was performed for further evaluation. Results: Our stent segmentation results were compared with the manually segmented contours. For the random‐split test, the trained model achieved a mean (±standard deviation) stent Dice similarity coefficient (DSC), 95% Hausdorff distance (HD95), the center‐of‐mass distance (CMD), and volume difference Voldiff$Vo{l_{diff}}$ were 0.96 (±0.01), 1.01 (±0.55) mm, 0.66 (±0.46) mm, and 3.07% (±2.37%), respectively. The sevenfold cross‐validation of the proposed PAUnet had the mean (±standard deviation) of 0.96 (±0.02), 0.72 (±0.49) mm, 0.85 (±0.96) mm, and 3.47% (±3.27%) for the DSC, HD95, CMD, and Voldiff$Vo{l_{diff}}$. Conclusion: We developed a novel deep learning–based approach to automatically segment the stent from IMR images, demonstrated its clinical feasibility, and validated its accuracy compared to manual segmentation. The proposed technique could be a useful tool for quantitative intrafraction motion monitoring in pancreatic cancer radiotherapy. [ABSTRACT FROM AUTHOR]