Deep Learning Architecture Research Articles

Diagnostic performance of artificial intelligence in detecting oral potentially malignant disorders and oral cancer using medical diagnostic imaging: a systematic review and meta-analysis

ObjectiveOral cancer is a widespread global health problem characterised by high mortality rates, wherein early detection is critical for better survival outcomes and quality of life. While visual examination is the primary method for detecting oral cancer, it may not be practical in remote areas. AI algorithms have shown some promise in detecting cancer from medical images, but their effectiveness in oral cancer detection remains Naïve. This systematic review aims to provide an extensive assessment of the existing evidence about the diagnostic accuracy of AI-driven approaches for detecting oral potentially malignant disorders (OPMDs) and oral cancer using medical diagnostic imaging.MethodsAdhering to PRISMA guidelines, the review scrutinised literature from PubMed, Scopus, and IEEE databases, with a specific focus on evaluating the performance of AI architectures across diverse imaging modalities for the detection of these conditions.ResultsThe performance of AI models, measured by sensitivity and specificity, was assessed using a hierarchical summary receiver operating characteristic (SROC) curve, with heterogeneity quantified through I2 statistic. To account for inter-study variability, a random effects model was utilized. We screened 296 articles, included 55 studies for qualitative synthesis, and selected 18 studies for meta-analysis. Studies evaluating the diagnostic efficacy of AI-based methods reveal a high sensitivity of 0.87 and specificity of 0.81. The diagnostic odds ratio (DOR) of 131.63 indicates a high likelihood of accurate diagnosis of oral cancer and OPMDs. The SROC curve (AUC) of 0.9758 indicates the exceptional diagnostic performance of such models. The research showed that deep learning (DL) architectures, especially CNNs (convolutional neural networks), were the best at finding OPMDs and oral cancer. Histopathological images exhibited the greatest sensitivity and specificity in these detections.ConclusionThese findings suggest that AI algorithms have the potential to function as reliable tools for the early diagnosis of OPMDs and oral cancer, offering significant advantages, particularly in resource-constrained settings.Systematic Review Registrationhttps://www.crd.york.ac.uk/, PROSPERO (CRD42023476706).

Correction to: Automatic liver segmentation using U-Net deep learning architecture for additive manufacturing

Correction to: Automatic liver segmentation using U-Net deep learning architecture for additive manufacturing

Global point cloud registration network for large transformations

Three-dimensional registration is an established yet challenging problem that is key in many different applications, such as mapping the environment for autonomous vehicles, or modeling people for avatar creation, among others. Registration refers to the process of mapping multiple data into the same coordinate system by means of matching correspondences and transformation estimation. Novel proposals exploit the benefits of deep learning architectures for this purpose, as they learn the best features for the data, providing better matches and hence results. However, the state of the art is usually focused on cases of relatively small transformations, although in certain applications and in a real and practical environment, large transformations are very common. In this paper, we present ReLaTo (Registration for Large Transformations), an architecture that addresses the cases where large transformations happen while maintaining good performance for local transformations. This proposal uses a novel Softmax pooling layer to find correspondences in a bilateral consensus manner between two point sets, sampling the most confident matches. These matches estimate a coarse and global registration using weighted Singular Value Decomposition (SVD). A target-guided denoising step is applied to both the obtained matches and latent features to estimate the final fine registration considering the local geometry. All these steps are carried out following an end-to-end approach, which has been shown to perform better than 7 state-of-the-art registration methods in two datasets commonly used for this task (ModelNet40 and the Karlsruhe Institute of Technology and Toyota Technological Institute dataset, KITTI), especially in the case of large transformations.Graphic abstract

A review of multimodal deep learning methods for genomic-enabled prediction in plant breeding.

Deep learning methods have been applied when working to enhance the prediction accuracy of traditional statistical methods in the field of plant breeding. Although deep learning seems to be a promising approach for genomic prediction, it has proven to have some limitations, since its conventional methods fail to leverage all available information. Multimodal deep learning methods aim to improve the predictive power of their unimodal counterparts by introducing several modalities (sources) of input information. In this review, we introduce some theoretical basic concepts of multimodal deep learning and provide a list of the most widely used neural network architectures in deep learning, as well as the available strategies to fuse data from different modalities. We mention some of the available computational resources for the practical implementation of multimodal deep learning problems. We finally performed a review of applications of multimodal deep learning to genomic selection in plant breeding and other related fields. We present a meta-picture of the practical performance of multimodal deep learning methods to highlight how these tools can help address complex problems in the field of plant breeding. We discussed some relevant considerations that researchers should keep in mind when applying multimodal deep learning methods. Multimodal deep learning holds significant potential for various fields, including genomic selection. While multimodal deep learning displays enhanced prediction capabilities over unimodal deep learning and other machine learning methods, it demands more computational resources. Multimodal deep learning effectively captures intermodal interactions, especially when integrating data from different sources. To apply multimodal deep learning in genomic selection, suitable architectures and fusion strategies must be chosen. It is relevant to keep in mind that multimodal deep learning, like unimodal deep learning, is a powerful tool but should be carefully applied. Given its predictive edge over traditional methods, multimodal deep learning is valuable in addressing challenges in plant breeding and food security amid a growing global population.

Prediction and clustering of Alzheimer’s disease by race and sex: a multi-head deep-learning approach to analyze irregular and heterogeneous data

Early detection of Alzheimer’s disease (AD) is crucial to maximize clinical outcomes. Most disease progression analyses include people with diagnoses of cognitive impairment, limiting understanding of AD risk among those with normal cognition. The objective was to establish AD progression models through a deep learning approach to analyze heterogeneous, multi-modal datasets, including clustering analyses of population subsets. A multi-head deep-learning architecture was built to process and learn from biomedical and imaging data from the National Alzheimer’s Coordinating Center. Shapley additive explanation algorithms for feature importance ranking and pairwise correlation analysis were used to identify predictors of disease progression. Four primary disease progression clusters (slow, moderate and rapid converters or non-converters) were subdivided into groups by race and sex, yielding 16 sub-clusters of participants with distinct progression patterns. A multi-head and early-fusion convolutional neural network achieved the most competitive performance and demonstrated superiority over a single-head deep learning architecture and conventional tree-based machine-learning methods, with 97% test accuracy, 96% F1 score and 0.19 root mean square error. From 447 features, 2 sets of 100 predictors of disease progression were extracted. Feature importance ranking, correlation analysis and descriptive statistics further enriched cluster analysis and validation of the heterogeneity of risk factors.

Deep learning for surgical instrument recognition and segmentation in robotic-assisted surgeries: a systematic review

Applying deep learning (DL) for annotating surgical instruments in robot-assisted minimally invasive surgeries (MIS) represents a significant advancement in surgical technology. This systematic review examines 48 studies that utilize advanced DL methods and architectures. These sophisticated DL models have shown notable improvements in the precision and efficiency of detecting and segmenting surgical tools. The enhanced capabilities of these models support various clinical applications, including real-time intraoperative guidance, comprehensive postoperative evaluations, and objective assessments of surgical skills. By accurately identifying and segmenting surgical instruments in video data, DL models provide detailed feedback to surgeons, thereby improving surgical outcomes and reducing complication risks. Furthermore, the application of DL in surgical education is transformative. The review underscores the significant impact of DL on improving the accuracy of skill assessments and the overall quality of surgical training programs. However, implementing DL in surgical tool detection and segmentation faces challenges, such as the need for large, accurately annotated datasets to train these models effectively. The manual annotation process is labor-intensive and time-consuming, posing a significant bottleneck. Future research should focus on automating the detection and segmentation process and enhancing the robustness of DL models against environmental variations. Expanding the application of DL models across various surgical specialties will be essential to fully realize this technology’s potential. Integrating DL with other emerging technologies, such as augmented reality (AR), also offers promising opportunities to further enhance the precision and efficacy of surgical procedures.

Deep-Sea Fauna Segmentation: A Comparative Analysis of Convolutional and Vision Transformer Architectures at Lucky Strike Vent Field

Abstract. Due to recent technological developments, the acquisition and availability of deep-sea imagery has increased exponentially in the last years, leading to an increasing backlog in image annotation and processing, attributable to limited specialized human resources. In this work, we investigate the performance of well-established convolutional neural networks and Vision Transformer (ViT) based architectures, namely, DeepLabv3+ and UNETR, for the segmentation of fauna in deep-sea images. The dataset consists of images captured at the Lucky Strike Vent field, located on the mid-Atlantic ridge, of three edifices named Montsegur, White Castle, and Eiffel Tower. Our experimental investigation reveals that the Vision Transformer consistently outperforms the fully convolutional deep learning architecture, by approximately 14% in terms of F1-Score, demonstrating the effectiveness of ViTs in capturing intricate patterns and long-range dependencies present in deep-sea imagery. Our findings highlight the potential of ViTs as a promising approach for accurate semantic segmentation in challenging environmental contexts, paving the way for improved understanding and analysis of deep-sea ecosystems.

A Deep Learning Framework with Learning without Forgetting for Intelligent Surveillance in IoT-enabled Home Environments in Smart Cities

Background: Internet of Things (IoT) technology in smart urban homes has revolutionised sophisticated monitoring. This progress uses interconnected devices and systems to improve security, resource management, and resident safety. Smart cities use technology to improve efficiency, sustainability, and quality. Internet of Things-enabled intelligent monitoring technologies are key to this goal. Objectives: Intelligent monitoring in IoT-enabled homes in smart cities improves security, convenience, and quality of life from advanced technologies. Using live monitoring and risk identification tools to quickly discover and resolve security breaches and suspicious activity to protect citizens. Intelligent devices allow homeowners to remotely control lighting, security locks, and surveillance cameras. Using advanced technologies to regulate heating, cooling, and lighting based on occupancy and usage. Method: This study introduces a deep learning architecture that uses LwF (Learning without Forgetting) to keep patterns while absorbing new data. The authors use IoT devices to collect and analyse data in real-time for monitoring and surveillance. They use sophisticated data preprocessing to handle IoT devices' massive data. The authors train the deep learning model with historical and real-time data and cross-validation to ensure resilience. Result: The proposed model has been validated on two different Robloflow datasets of 7382 images. The proposed model gains an accuracy level of 98.27%. The proposed Yolo-LwF model outperforms both the original Yolo and LwF models in terms of detection speed and adaptive learning. Conclusion: By raising the bar for intelligent monitoring solutions in smart cities, the suggested system is ideal for real-time, adaptive surveillance in IoT-enabled households. By embracing adaptability and knowledge retention, authors envision heightened security and safety levels in urban settings.

Evaluating Uncertainty Quantification in Medical Image Segmentation: A Multi-Dataset, Multi-Algorithm Study

Deep learning is revolutionizing various scientific fields, with medical applications at the forefront. One key focus is automating image segmentation, a process crucial in many clinical services. However, medical images are often ambiguous and challenging even for experts. To address this, reliable models need to quantify their uncertainty, allowing physicians to understand the model’s confidence in its segmentation. This paper explores how the performance and uncertainty of a model are influenced by the number of annotations per input sample. We examine the effects of both single and multiple manual annotations on various deep learning architectures. To tackle this question, we employ three widely recognized deep learning architectures and evaluate them across four publicly available datasets. Furthermore, we explore the effects of dropout rates on Monte Carlo models by examining uncertainty models with dropout rates of 20%, 40%, 60%, and 80%. Subsequently, we evaluate the models using various measurement metrics. The findings reveal that the influence of multiple annotations varies significantly depending on the datasets. Additionally, we observe that the dropout rate has minimal or no impact on the model’s performance unless there is a substantial loss of training data, primarily evident in the 80% dropout rate scenario.

Nested Neural Networks: A Novel Approach to Flexible and Deep Learning Architectures

It’s easy to get lost in the complexities of sentences like, “Never will I have been in a place whose experiences will have been teaching me how amazing life really is till the beginning of every one of those marvelous whole new advenchallenges for more than 20 years of life by the end of my military service.” These kinds of nested and layered statements are not just the domain of advanced grammar—they have a parallel in cutting-edge technology as well. This paper introduces a novel neural network architecture, termed Nested Neural Networks (NNN), that incorporates nested layers within a more complex overarching structure. Much like the layered conditions in Future Perfect and Future Perfect Progressive tenses, NNNs utilize ‘nested’ conditions to bring flexibility and depth to data processing. This structure captures complex relationships, achieving high performance, computational efficiency, and generalization across tasks, much as understanding the nested clauses of a sentence can reveal deeper meanings. Our experiments on benchmark datasets demonstrate NNNs’ ability to handle intricate data patterns with enhanced memory efficiency and training adaptability, often outperforming traditional models. This fascinating intersection between language and machine learning showcases the powerful potential of structured complexity, whether in communication or computation.

Price spread prediction in high-frequency pairs trading using deep learning architectures

Price spread prediction in high-frequency pairs trading using deep learning architectures

A two-stage deep learning architecture for detection global coastal and offshore submesoscale ocean eddy using SDGSAT-1 multispectral imagery

Submesoscale ocean eddies are essential oceanic phenomenon that control and influence the ocean energy cascade. Most existing eddy detection methods rely on low-resolution satellite altimeter data, which fail to capture submesoscale ocean features and oceanographic phenomena in shallow water. Introducing high-resolution multispectral data can alleviate these problems, yet it has been largely overlooked. A generalized and efficient deep learning architecture that combines developments in deep learning with Sustainable Development Goals Science Satellite 1 (SDGSAT-1) multispectral data from earth observations offers a potential pathway for more fine detection of ocean eddies. Considering that oceanic eddy exhibits spatially sparse characteristics on high-resolution remote sensing scenes, the oceanic eddy detection (OED) model suitable for global coastal and offshore regions is divided into two stages: eddy information judgment and eddy position determination. Correspondingly, SDGSAT-1 multispectral data from November 2021 to December 2022 were carried out to construct two submesoscale eddy datasets for training and testing each stage model. The union validation of multiple metrics demonstrates that the proposed OED model and its stage models achieve state-of-the-art (SOTA) performance, especially in optically complex coastal and offshore waters. We applied the model to real-world scenes captured by SDGSAT-1 in 2023, and found that the detected results were mainly located at the water depth below 200 m. The authenticity of the recognition results is validated using sea surface chlorophyll concentration, temperature, and topography data, indicating that the OED model has achieved remarkable effectiveness under various sea conditions. In addition, the temporal distributions and statistical characteristics of detected submesoscale eddies are analyzed over an extended period (November 2021 to November 2023). Finally, HISEA-2, Landsat-9, and Sentinel-2 served as testing grounds to validate the generalization of the proposed methodology, with experimental results demonstrating that the OED model possesses significant developmental potential for multi-source remote sensing data. This paper presents a comprehensive deep learning framework for the global-scale detection of submesoscale eddies and underscores the pivotal role of high-resolution multispectral imagery as an innovative data source for global coastal and offshore eddy identification.

Multi-sensor Information Fusion in Internet of Vehicles Based on Deep Learning: A Review

Environmental perception is a crucial component of intelligent driving technology, providing the informational foundation for intelligent decision-making and collaborative control. Due to the limitations of single sensors and the continuous advancements in deep learning and sensor technologies, multi-sensor information fusion in the Internet of Vehicles (IoV) has emerged as a major research hotspot. This approach is also a primary solution for achieving full self-driving. However, given the complexity of the technology, there are still many challenges in achieving accurate and reliable real-time multi-source information perception. Current discussions often focus on specific aspects of multi-sensor fusion in intelligent driving, while detailed discussions on sensor fusion in the context of the IoV are relatively scarce. To provide a comprehensive discussion and analysis of multi-sensor information fusion in IoV, this paper first provides a detailed introduction to its developmental background and the commonly involved sensors. Subsequently, a detailed analysis of the strategies, deep learning architectures, and methods for multi-sensor information fusion in the IoV is presented. Finally, the specific applications and key issues related to multi-sensor information fusion in IoV are discussed from multiple perspectives, along with an analysis of future development trends. This paper aims to serve as a valuable reference for advancing multi-sensor information fusion technology in IoV environments and supporting the realization of full self-driving.

Investigating neural networks with groundwater flow equation loss

Physics-Informed Neural Networks (PINNs) are considered a powerful tool for solving partial differential equations (PDEs), particularly for the groundwater flow (GF) problem. In this paper, we investigate how the deep learning (DL) architecture, within the PINN framework, is connected to the ability to compute a more or less accurate numerical GF solution, so the link ‘PINN architecture - numerical performance’ is explored. Specifically, this paper explores the effect of various DL components, such as different activation functions and neural network structures, on the computational framework. Through numerical results and on the basis of some theoretical foundations of PINNs, this research aims to improve the explicability of PINNs to resolve, in this case, the one-dimensional GF equation. Moreover, our problem involves source terms described by a Dirac delta function, providing insights into the role of DL architecture in solving complex PDEs.

High-resolution acoustic-impedance inversion based on a deep-learning-aided representation model of nonstationary seismic data

Building a high-resolution acoustic-impedance (AI) model based on nonstationary seismic data plays a key role in reservoir predictions. However, the common AI inversion methods are faced with two main shortcomings. First, because of the nonstationary feature of seismic data, a multiparameter inverse problem should be considered to not only estimate AI but also to estimate a time-varying wavelet or a quality factor ( Q) model, leading the problem to be more ill posed. Second, the resolution of the estimated AI is limited due to the band-limited nature of nonstationary seismic data. To address these issues, we develop a new nonstationary seismic AI inversion method. Notably, we develop a deep-learning-aided representation model to replace the nonstationary convolution model to solve the forward problem in inversion. Benefiting from multisource information (seismic data and well-log data) and the powerful nonlinear function fitting ability of deep learning, this model can map high-resolution AI to band-limited nonstationary seismic data without requiring a time-varying wavelet or a Q model. A new deep-learning architecture is developed for processing the spatio-temporal seismic data with better accuracy. In addition, total-variation regularization is adopted to enforce a physically reasonable AI model. The results of our 3D synthetic and field data experiments clearly demonstrate that our method has significant advantages over other common methods in building a high-resolution AI model.

Channel reconstruction through improvised deep learning architecture for high-speed networks

Efficient acquisition of channel state information (CSI) is quite complicated process but immensely essential to exploit probable benefits of massive multiple input multiple output (MIMO) systems. Therefore, a deep learningbased model is utilized to estimate channel feedback in a massive MIMO system. The proposed improvised deep learning-based channel estimation (IDLCE) model enhances channel reconstruction efficiency by using multiple convolutional layers and residual blocks. The proposed IDLCE model utilizes encoder network to compress CSI matrices where decoder network is used to downlink reconstruct CSI matrices. Here, an additional quantization block is incorporated to improve feedback reconstruction accuracy by reducing channel errors. A COST 2,100 model is adopted to analyse performance efficiency for both indoor and outdoor scenarios. Further, deep learning-based model is used to train thousands of parameter and correlation coefficients much faster and to minimize computational complexity. The proposed IDLCE model evaluate performance in terms of normalized mean square error (NMSE), correlation efficiency and reconstruction accuracy and compared against varied state-of-art-channel estimation techniques. Excellent performance results are obtained with large improvement in channel reconstruction accuracy

HarDNet-based deep learning model for osteoporosis screening and bone mineral density inference from hand radiographs

PurposeOsteoporosis, affecting over 200 million individuals, often remains unrecognized and untreated, increasing the risk of fractures in older adults. Osteoporosis is typically diagnosed with bone mineral density (BMD) measured by dual-energy X-ray absorptiometry (DXA). This study aims to develop DeepDXA-Hand, a deep learning model using the efficient CNN-based deep learning architecture, for opportunistic osteoporosis screening from hand radiographs. MethodsDeepDXA-Hand utilizes a CNN-based, HarDNet, approach to predict BMD non-invasively. A total of 10,351 hand radiographs and DXA pairs were used for model training and validation. The model's interpretability was enhanced using GradCAM for hotspot analysis to determine the model's attention areas. ResultsThe predicted and ground truth BMD were significantly correlated with a correlation coefficient of 0.745. For binary classification of osteoporosis, DeepDXA-Hand demonstrated a sensitivity of 0.73, specificity of 0.83, and accuracy of 0.80, indicating its clinical potential. The model mainly focused on the carpal bones, such as the capitate, trapezoid, hamate, triquetrum, and the head of the second metacarpal bone, suggesting these areas provide radiological features for inferring BMD. ConclusionDeepDXA-Hand shows potential for the early detection of osteoporosis with high sensitivity and specificity. Further studies should explore its utility in predicting fracture risks. Mini abstractOsteoporosis affects millions and often goes undetected and untreated. DeepDXA-Hand, a HarDNet-based deep learning model, predicted bone mineral density with a correlation of 0.745 and classified osteoporosis with 0.80 accuracy. This model enhances early detection and has significant clinical potential as osteoporosis opportunistic screening tool.

Enhancing performance of end-to-end communication system using Attention Mechanism-based Sparse Autoencoder over Rayleigh fading channel

Deep learning has revolutionized communication systems by introducing innovative approaches to address channel impairments through end-to-end models. Autoencoders, a type of deep learning architecture, are adept at learning compact data representations. However, conventional autoencoders in end-to-end models can suffer from overfitting, which limits their effectiveness in noisy communication environments. To address this issue, we propose a Sparse Autoencoder-based (SAE) model that enforces sparsity and promotes the extraction of robust features. Despite its effectiveness, the SAE model may still lack the ability to focus on the most relevant features of the input data. To overcome this limitation, we further introduce an Attention Mechanism-based Sparse Autoencoder (ASA) model. This model integrates the feature extraction capabilities of a sparse autoencoder with an attention mechanism that selectively highlights informative features of the signal. Through simulations, we demonstrate that both proposed models significantly improve M-PSK and M-QAM communication system performance. When trained at 7 dB, both proposed models exhibit significant performance improvements at higher testing average SNRs. Our results show that the SAE model outperforms the conventional Maximum Likelihood Detection (MLD) model and baseline autoencoder systems but suffers from error floor issues. The SAE model suffers from an error floor at average SNRs beyond 16 dB for BPSK and 14 dB for higher-order modulation schemes. As the value of M increases, the performance gap between the MLD and the proposed SAE model narrows. The ASA model, however, effectively mitigates the error floor observed in the SAE model for all values of M and across all modulation schemes. This research highlights the benefits of integrating an attention mechanism with SAE, resulting in enhanced robustness and reliability in communication systems characterized by improved accuracy and reduced error rates.

DSCIMABNet: A novel multi-head attention depthwise separable CNN model for skin cancer detection

Skin cancer is a common type of cancer worldwide. Early diagnosis of skin cancer can reduce the risk of death by increasing treatment success. However, it is challenging for dermatologists or specialists because the symptoms are vague in the early stages and cannot be noticed by the naked eye. This study examines digital diagnostic techniques supported by artificial intelligence, focusing on early skin cancer detection and two methods have been proposed. In the first method, DSCIMABNet deep learning architecture was developed by combining multi-head attention and depthwise separable convolution techniques. This model provides flexibility in learning the dataset's local features, abstract concepts, and long-term relationships. The DSCIMABNet model and modern deep learning models trained on ImageNet are proposed to be combined with the ensemble learning method in the second method. This approach provides a comprehensive feature extraction process that will increase the performance of the classification process with ensemble learning. The proposed approaches are trained and evaluated on the ISIC 2018 dataset with image enhancement applied in preprocessing. In the experimental results, DSCIMABNet achieved 84.28% accuracy, while the proposed hybrid method achieved 99.40% accuracy. Moreover, on the Mendeley dataset (CNN for Melanoma Detection Data), DSCIMABNet achieved 92.58% accuracy, while the hybrid method achieved 99.37% accuracy. This study may significantly contribute to developing new and effective methods for the early diagnosis and treatment of skin cancer.

Thoracic Aortic Three-Dimensional Geometry.

Aortic structure impacts cardiovascular health through multiple mechanisms. Aortic structural degeneration occurs with aging, increasing left ventricular afterload and promoting increased arterial pulsatility and target organ damage. Despite the impact of aortic structure on cardiovascular health, three-dimensional (3D) aortic geometry has not been comprehensively characterized in large populations. We segmented the complete thoracic aorta using a deep learning architecture and used morphological image operations to extract multiple aortic geometric phenotypes (AGPs, including diameter, length, curvature, and tortuosity) across various subsegments of the thoracic aorta. We deployed our segmentation approach on imaging scans from 54,241 participants in the UK Biobank and 8,456 participants in the Penn Medicine Biobank. Our method provides a fully automated approach towards quantifying the three-dimensional structural parameters of the aorta. This approach expands the available phenotypes in two large representative biobanks and will allow large-scale studies to elucidate the biology and clinical consequences of aortic degeneration related to aging and disease states.