Deep Learning Methods Research Articles

Molecular mechanism underlying effect of D93 and D289 protonation states on inhibitor-BACE1 binding: exploration from multiple independent Gaussian accelerated molecular dynamics and deep learning

ABSTRACT BACE1 has been regarded as an essential drug design target for treating Alzheimer’s disease (AD). Multiple independent Gaussian accelerated molecular dynamics simulations (GaMD), deep learning (DL), and molecular mechanics general Born surface area (MM-GBSA) method are integrated to elucidate the molecular mechanism underlying the effect of D93 and D289 protonation on binding of inhibitors OV6 and 4B2 to BACE1. The GaMD trajectory-based DL successfully identifies significant function domains. Dynamic analysis shows that the protonation of D93 and D289 strongly affects the structural flexibility and dynamic behaviour of BACE1. Free energy landscapes indicate that inhibitor-bound BACE1s have more conformational states in the protonated states than the wild-type (WT) BACE1, and show more binding poses of inhibitors. Binding affinities calculated using the MM-GBSA method indicate that the protonation of D93 and D289 highly disturbs the binding ability of inhibitors to BACE1. In addition, the protonation of two residues significantly affects the hydrogen bonding interactions (HBIs) of OV6 and 4B2 with BACE1, altering their binding activity to BACE1. The binding hot spots of BACE1 recognized by residue-based free energy estimations provide rational targeting sites for drug design towards BACE1. This study is anticipated to provide theoretical aids for drug development towards treatment of AD.

Decision-Making Policy for Autonomous Vehicles on Highways Using Deep Reinforcement Learning (DRL) Method

Automated driving (AD) is a new technology that aims to mitigate traffic accidents and enhance driving efficiency. This study presents a deep reinforcement learning (DRL) method for autonomous vehicles that can safely and efficiently handle highway overtaking scenarios. The first step is to create a highway traffic environment where the agent can be guided safely through surrounding vehicles. A hierarchical control framework is then provided to manage high-level driving decisions and low-level control commands, such as speed and acceleration. Next, a special DRL-based method called deep deterministic policy gradient (DDPG) is used to derive decision strategies for use on the highway. The performance of the DDPG algorithm is compared with that of the DQN and PPO algorithms, and the results are evaluated. The simulation results show that the DDPG algorithm can effectively and safely handle highway traffic tasks.

PET/CT radiomics and deep learning in the diagnosis of benign and malignant pulmonary nodules: progress and challenges

Lung cancer is currently the leading cause of cancer-related deaths, and early diagnosis and screening can significantly reduce its mortality rate. Since some early-stage lung cancers lack obvious clinical symptoms and only present as pulmonary nodules (PNs) in imaging examinations, accurately determining the benign or malignant nature of PNs is crucial for improving patient survival rates. 18F-FDG PET/CT is important in diagnosing PNs, but its specificity needs improvement. Radiomics can provide information beyond traditional visual assessment, overcoming its limitations by extracting high-throughput quantitative features from medical images. Radiomics features based on 18F-FDG PET/CT and deep learning methods have shown great potential in the noninvasive diagnosis of PNs. This paper reviews the latest advancements in these methods and discusses their contributions to improving diagnostic accuracy and the challenges they face.

GPS-pPLM: A Language Model for Prediction of Prokaryotic Phosphorylation Sites

In the prokaryotic kingdom, protein phosphorylation serves as one of the most important posttranslational modifications (PTMs) and is involved in orchestrating a broad spectrum of biological processes. Here, we report an updated online server named the group-based prediction system for prokaryotic phosphorylation language model (GPS-pPLM), used for predicting phosphorylation sites (p-sites) in prokaryotes. For model training, two deep learning methods, a transformer and a deep neural network, were employed, and a total of 10 sequence features and contextual features were integrated. Using 44,839 nonredundant p-sites in 16,041 proteins from 95 prokaryotes, two general models for the prediction of O-phosphorylation and N-phosphorylation were first pretrained and then fine-tuned to construct 6 predictors specific for each phosphorylatable residue type as well as 134 species-specific predictors. Compared with other existing tools, the GPS-pPLM exhibits higher accuracy in predicting prokaryotic O-phosphorylation p-sites. Protein sequences in FASTA format or UniProt accession numbers can be submitted by users, and the predicted results are displayed in tabular form. In addition, we annotate the predicted p-sites with knowledge from 22 public resources, including experimental evidence, 3D structures, and disorder tendencies. The online service of the GPS-pPLM is freely accessible for academic research.

Detection of road extraction from satellite images with deep learning method

Detection of road extraction from satellite images with deep learning method

A Comprehensive Review of Mobile Robot Navigation Using Deep Reinforcement Learning Algorithms in Crowded Environments

Navigation is a crucial challenge for mobile robots. Currently, deep reinforcement learning has attracted considerable attention and has witnessed substantial development owing to its robust performance and learning capabilities in real-world scenarios. Scientists leverage the advantages of deep neural networks, such as long short-term memory, recurrent neural networks, and convolutional neural networks, to integrate them into mobile robot navigation based on deep reinforcement learning. This integration aims to enhance the robot's motion control performance in both static and dynamic environments. This paper illustrates a comprehensive survey of deep reinforcement learning methods applied to mobile robot navigation systems in crowded environments, exploring various navigation frameworks based on deep reinforcement learning and their benefits over traditional simultaneous localization and mapping-based frameworks. Subsequently, we comprehensively compare and analyze the relationships and differences among three types of navigation: autonomous-based navigation, navigation based on simultaneous localization and mapping, and planning-based navigation. Moreover, the crowded environment includes static, dynamic, and a combination of obstacles in different typical application scenarios. Finally, we offer insights into the evolution of navigation based on deep reinforcement learning, addressing the problems and providing potential solutions associated with this emerging field.

Deep learning method for detecting fluorescence spots in cancer diagnostics via fluorescence in situ hybridization

Fluorescence in Situ Hybridization (FISH) is a technique for macromolecule identification that utilizes the complementarity of DNA or DNA/RNA double strands. Probes, crafted from selected DNA strands tagged with fluorophore-coupled nucleotides, hybridize to complementary sequences within the cells and tissues under examination. These are subsequently visualized through fluorescence microscopy or imaging systems. However, the vast number of cells and disorganized nucleic acid sequences in FISH images present significant challenges. The manual processing and analysis of these images are not only time-consuming but also prone to human error due to visual fatigue. To overcome these challenges, we propose the integration of medical imaging with deep learning to develop an automated detection system for FISH images. This system features an algorithm capable of quickly detecting fluorescent spots and capturing their coordinates, which is crucial for evaluating cellular characteristics in cancer diagnosis. Traditional models struggle with the small size, low resolution, and noise prevalent in fluorescent points, leading to significant performance declines. This paper offers a detailed examination of these issues, providing insights into why traditional models falter. Comparative tests between the YOLO series models and our proposed method affirm the superior accuracy of our approach in identifying fluorescent dots in FISH images.

Advances in ECG and PCG-based cardiovascular disease classification: a review of deep learning and machine learning methods

Cardiovascular diseases (CVD) have been found to be prevalent in society, frequently ending in death. According to the findings of a recent survey, the mortality rate is increasing due to the prevalence of adult cigarette consumption, elevated blood pressure, high cholesterol levels, and obesity. The previously mentioned causes are exacerbating the severity of the condition. A pressing necessity exists for a study on the variability of these factors and their impact on cardiovascular disease (CVD). This involves the use of advanced tools to detect the disease early on and aid in the reduction of fatality rates. With their extensive methodologies that would help in the early CVD prediction and recognition of behavioral patterns in large amounts of data, artificial intelligence, and data mining disciplines offer a broad study potential. The results of these predictions will help physicians make decisions and early diagnoses, decreasing the risk of patient death. This work compares and reports the classification, machine learning, and deep learning algorithms that predict cardiovascular illnesses. For this study, articles from 2012 to 2023 were considered; after filtering, 82 articles were chosen for primary research. Future researchers will benefit from this review on cardiovascular disorders by better understanding the Deep Learning and Machine Learning models now in the healthcare sector. The review encompasses commonly employed methodologies such as support vector machine, decision tree, random forest, and convolutional neural networks (CNNs). Additionally, this survey aggregates and presents information on the performance metrics used to report accuracy. It also goes over the most popular datasets used by various diagnostic models (ECG and PCG signals datasets). In addition, it emphasizes prominent publishers, journals, and conferences that serve as platforms for the evaluation of scholarly works. Additionally, it will facilitate their understanding of the unresolved challenges or hurdles experienced by past researchers. A lack of more extensive and consistent datasets was the most common issue, followed by the need to improve existing models.

TPTrans: Vessel Trajectory Prediction Model Based on Transformer Using AIS Data

The analysis of large amounts of vessel trajectory data can facilitate more complex traffic management and route planning, thereby reducing the risk of accidents. The application of deep learning methods in vessel trajectory prediction is becoming more and more widespread; however, due to the complexity of the marine environment, including the influence of geographical environmental factors, weather factors, and real-time traffic conditions, predicting trajectories in less constrained maritime areas is more challenging than in path network conditions. Ship trajectory prediction methods based on kinematic formulas work well in ideal conditions but struggle with real-world complexities. Machine learning methods avoid kinematic formulas but fail to fully leverage complex data due to their simple structure. Deep learning methods, which do not require preset formulas, still face challenges in achieving high-precision and long-term predictions, particularly with complex ship movements and heterogeneous data. This study introduces an innovative model based on the transformer structure to predict the trajectory of a vessel. First, by processing the raw AIS (Automatic Identification System) data, we provide the model with a more efficient input format and data that are both more representative and concise. Secondly, we combine convolutional layers with the transformer structure, using convolutional neural networks to extract local spatiotemporal features in sequences. The encoder and decoder structure of the traditional transformer structure is retained by us. The attention mechanism is used to extract the global spatiotemporal features of sequences. Finally, the model is trained and tested using publicly available AIS data. The prediction results on the field data show that the model can predict trajectories including straight lines and turns under the field data of complex terrain, and in terms of prediction accuracy, our model can reduce the mean squared error by at least 6×10−4 compared with the baseline model.

Evaluation of coarse aggregate properties in hardened concrete based on segment anything model (SAM)

The applicability of the segment anything model (SAM) algorithm, a novel technique for analyzing the characteristics of coarse aggregate from RGB images of concrete cross-sections, was evaluated. Unlike traditional deep learning method, the SAM algorithm effectively and clearly separates coarse aggregate and mortar without requiring a large database, extensive pre-training, conversion of RGB images to grayscale, or manual threshold setting. Additionally, a new technique was applied to separate the pores recognized as aggregates by using the image difference according to the change in vertical and side lighting. Consequently, the area ratio of coarse aggregate, after the separation and removal of pores from the aggregate area, exhibited an error of approximately 7.0 % when compared to the actual volume ratio of the aggregate. However, in contrast to the calculation of the area ratio of coarse aggregate, the SAM algorithm demonstrated significant errors in size analysis, leading to distortions in the sieve curve. Nevertheless, the SAM algorithm indicates potential for application in various concrete-related fields of concrete in the future.

Assessing polyomic risk to predict Alzheimer's disease using a machine learning model

AbstractINTRODUCTIONAlzheimer's disease (AD) is the most common form of dementia in the elderly. Given that AD neuropathology begins decades before symptoms, there is a dire need for effective screening tools for early detection of AD to facilitate early intervention.METHODSHere, we used tree‐based and deep learning methods to train polyomic prediction models for AD affection status and age at onset, employing genomic, proteomic, metabolomic, and drug use data from UK Biobank. We used SHAP to determine the feature's importance.RESULTSOur best‐performing polyomic model achieved an area under the receiver operating characteristics curve (AUROC) of 0.87. We identified GFAP and CXCL17 proteins to be the strongest predictors of AD, besides apolipoprotein E (APOE) alleles. Increasing the number of cases by including “AD‐by‐proxy” cases did not improve AD prediction.DISCUSSIONAmong the four modalities, genomics, and proteomics were the most informative modality based on AUROC (area under the receiver operating characteristic curve). Our data suggest that two blood‐based biomarkers (glial fibrillary acidic protein [GFAP] and CXCL17) may be effective for early presymptomatic prediction of AD.Highlights We developed a polyomic model to predict AD and age‐at‐onset using omics and medication use data from EHR. We identified GFAP and CXCL17 proteins to be the strongest predictors of AD, besides APOE alleles. “AD‐by‐proxy” cases, if used in training, do not improve AD prediction. Proteomics was the most informative modality overall for affection status and AAO prediction.

MFHARFNet: multi-branch feature hybrid and adaptive receptive field network for image segmentation

Abstract Accurate segmentation of medical images is crucial for disease diagnosis and understanding disease changes. Deep learning methods, utilizing encoder-decoder structures, have demonstrated cutting-edge performance in various medical image segmentation tasks. However, the pooling operation in the encoding stage results in feature loss, which makes the network lack the ability to fuse multi-scale information at different levels, hinders its effective perception of multi-scale information, and leads to poor segmentation performance. Drawing inspiration from the U-shaped network, this study introduces a multi-branch feature hybrid attention and adaptive receptive field network (MFHARFNet) for medical image segmentation. Building upon the encoder-decoder framework, we initially devise a multi-branch feature hybrid attention module (MFHAM) to seamlessly integrate feature maps of varying scales, capturing both fine-grained features and coarse-grained semantics across the entire scale. Furthermore, we redesign the skip connection to amalgamate feature information from different branches in the encoder stage and efficiently transmit it to the decoder, providing the decoder with global context feature maps at different levels. Finally, the adaptive receptive field (ARF) module is introduced in the decoder feature reconstruction stage to adapt and focus on related fields, ensuring the model’s adaptation to different segmentation target features, and achieving different weights for the output of different convolution kernels to improve segmentation performance. We comprehensively evaluate our method on medical image segmentation tasks, by using four public datasets across CT and MRI. Remarkably, MFHARFNet method consistently outperforms other state-of-the-art methods, exceeding UNet by 2.1%, 0.9%, 6.6% and 1.0% on Dice on ATLAS, LiTs, BraTs2019 and Spine and intervertebral disc datasets, respectively. In addition, MFHARFNet minimizes network parameters and computational complexity as much as possible. The source codes are in https://github.com/OneHundred99/MFHARFNet.

Intelligent rehabilitation assistant: Application of deep learning methods in sports injury recovery

In recent years, sports injury rehabilitation has developed into a specialized field that has forced the combination of an orthopedic surgeon, sports physiotherapist, and sports physician. Determining the appropriate time for an injured athlete to resume practice or competition is regarded as sports rehabilitation. Discovering the best solutions to avoid injuries, maximize recovery, and enhance performance is crucial for sports activities. The study introduced an intelligent rehabilitation assistant (IRA) that leverages advanced deep learning (DL) methods to enhance sports injury recovery. In this study, the IRA incorporates redefined prairie dog optimized bidirectional long-short-term memory (RPDO-Bi-LSTM) to enhance accuracy, predicting sports injury recovery. The study collected data on the state of rehabilitation, physiological parameters, and general health using wearable sensors and movement patterns. The data was preprocessed using a median filter to remove noise from sensor data. Region-based segmentation using segmented images from preprocessed data. Convolutional neural networks (CNN) using extracted features from obtained data. The IRA provides personalized recovery plans and real-time feedback. The framework consists of the components, suggested models to create quality scores for motions, measurements to quantify motion performance, and scoring of performance measurement elements into numerical quality scores. The proposed method is implemented using Python software. RPDO-Bi-LSTM presentation is evaluated by various metrics, such as accuracy 94.2% recall 98.2%, precision 96.5%, and specificity 95.2%, f1 score 95.6%, he planned technique attained good performance and improved the accuracy of sports injury recovery.

Vision transformers for glioma classification using T1 magnetic resonance imaging

Automated image analysis and classification have increasingly advanced in recent decades owing to machine learning and computer vision. In particular, deep learning (DL) architectures have become popular in resource-limited and labor-restricted environments such as the health-care sector. Transformer architecture, a DL method with self-attention mechanism, excels in natural language processing; however, its application in image-based diagnosis in health-care sector remains limited. Herein, the feasibility, bottlenecks, and performance of transformers in magnetic resonance imaging (MRI)-based brain tumor classification were investigated. To this end, a vision transformer (ViT) model was trained and tested using the popular Brain Tumor Segmentation (BraTS) 2015 dataset for glioma classification. Owing to limited data availability, domain adaptation techniques were used to pretrain the ViT model and the BraTS 2015 dataset was used for its fine-tuning. With the model only trained for 100 epochs, the confusion matrix for the two-class problem of tumor and nontumor classification showed an overall classification accuracy of 81.8%. In conclusion, although convolutional neural networks are traditionally used for DL-based medical image classification owing to their attention mechanism and long-range dependency-capturing capability, ViTs can outperform them in MRI-based brain tumor classification.

Optimising Bearing Degradation Prediction: A Bayesian Vector Autoregressive-Assisted DeepAR Model Approach

Abstract Accurate prediction of the remaining useful life (RUL) of bearings is crucial for the maintenance of rotating machinery. Current research often involves extracting multidimensional features from vibration signals to construct health degradation indicator, and health indicator is directly used for degradation prediction. However, constructed health indicator exhibit complex nonlinearity and fail to utilize the correlations with the original multidimensional features. To address this limitation, we propose an integrated prediction framework, known as the Bayesian Vector Autoregression-assisted DeepAR method. The proposed two step prediction framework combines the advantages of covariance prediction and temporal correlation prediction, and realizes the full utilization of multidimensional features correlation information. Additionally, the proposed approach utilizes the intrinsic properties of statistical models to constrain improve the prediction results of deep learning methods, achieving nonlinear modeling of the degradation indicator while maintaining the stability of the prediction results. This process of prediction, re-estimation prediction and adaptive adjustment achieves a balance in model performance. Finally, experimental studies using the PRONOSTIA and XJTU-SY datasets validate the effectiveness of the proposed method in improving the prediction accuracy of bearing health degradation indicators and enhancing the stability of multistep prediction in deep model.

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.

Quadratic Unet for seismic random noise attenuation

In the field of seismic data processing, seismic denoising is essential to improve the quality of seismic data. Various deep learning methods have been proposed, showing promising performance for seismic denoising. However, most deep learning methods are based on lin ear neurons, resulting in limited expressive ability for complex seismic signals. To enhance denoising capability using non-linear neurons, we propose a supervised quadratic U-shaped network (QUnet) for seismic random noise attenuation. The quadratic neurons in QUnet are represented by a second-order polynomial of input data and weighted parameters. Nu merical synthetic seismic data experiments show that the proposed QUnet method achieves higher signal-to-noise ratio results and preserves the continuity of reflectors more effectively. We have also tested the effectiveness of QUnet on both prestack field data and post-stack seismic data. The detailed structures and weak signals of seismic data are better preserved by our proposed QUnet method.

Large-Kernel Central Block Masked Convolution and Channel Attention-Based Reconstruction Network for Anomaly Detection of High-Resolution Hyperspectral Imagery

In recent years, the rapid advancement of drone technology has led to an increasing use of drones equipped with hyperspectral sensors for ground imaging. Hyperspectral data captured via drones offer significantly higher spatial resolution, but this also introduces more complex background details and larger target scales in high-resolution hyperspectral imagery (HRHSI), posing substantial challenges for hyperspectral anomaly detection (HAD). Mainstream reconstruction-based deep learning methods predominantly emphasize spatial local information in hyperspectral images (HSIs), relying on small spatial neighborhoods for reconstruction. As a result, large anomalous targets and background details are often well reconstructed, leading to poor anomaly detection performance, as these targets are not sufficiently distinguished from the background. To address these limitations, we propose a novel HAD network for HRHSI based on large-kernel central block masked convolution and channel attention, termed LKCMCA. Specifically, we first employ the pixel-shuffle technique to reduce the size of anomalous targets without losing image information. Next, we design a large-kernel central block masked convolution to make the network pay more attention to the surrounding background information, enabling better fusion of the information between adjacent bands. This, coupled with an efficient channel attention mechanism, allows the network to capture deeper spectral features, enhancing the reconstruction of the background while suppressing anomalous targets. Furthermore, we introduce an adaptive loss function by down-weighting anomalous pixels based on the mean absolute error. This loss function is specifically designed to suppress the reconstruction of potentially anomalous pixels during network training, allowing our model to be considered an excellent background reconstruction network. By leveraging reconstruction error, the model effectively highlights anomalous targets. Meanwhile, we produced four benchmark datasets specifically for HAD tasks using existing HRHSI data, addressing the current shortage of HRHSI datasets in the HAD field. Extensive experiments demonstrate that our LKCMCA method achieves superior detection performance, outperforming ten state-of-the-art HAD methods on all datasets.

Classification of EEG evoked in 2D and 3D virtual reality: traditional machine learning versus deep learning

Backgrounds. Virtual reality (VR) simulates real-life events and scenarios and is widely utilized in education, entertainment, and medicine. VR can be presented in two dimensions (2D) or three dimensions (3D), with 3D VR offering a more realistic and immersive experience. Previous research has shown that electroencephalogram (EEG) profiles induced by 3D VR differ from those of 2D VR in various aspects, including brain rhythm power, activation, and functional connectivity. However, studies focused on classifying EEG in 2D and 3D VR contexts remain limited.Methods. A 56-channel EEG was recorded while visual stimuli were presented in 2D and 3D VR. The recorded EEG signals were classified using two machine learning approaches: traditional machine learning and deep learning. In the traditional approach, features such as power spectral density (PSD) and common spatial patterns (CSP) were extracted, and three classifiers-support vector machines (SVM), K-nearest neighbors (KNN), and random forests (RF)-were used. For the deep learning approach, a specialized convolutional neural network, EEGNet, was employed. The classification performance of these methods was then compared.Results. In terms of accuracy, precision, recall, and F1-score, the deep learning method outperformed traditional machine learning approaches. Specifically, the classification accuracy using the EEGNet deep learning model reached up to 97.86%.Conclusions. EEGNet-based deep learning significantly outperforms conventional machine learning methods in classifying EEG signals induced by 2D and 3D VR. Given EEGNet's design for EEG-based brain-computer interfaces (BCI), this superior classification performance suggests that it can enhance the application of 3D VR in BCI systems.

Three-dimensional localization and tracking of chromosomal loci throughout the Escherichia coli cell cycle.

The intracellular position of genes may impact their expression, but it has not been possible to accurately measure the 3D position of chromosomal loci. In 2D, loci can be tracked using arrays of DNA-binding sites for transcription factors (TFs) fused with fluorescent proteins. However, the same 2D data can result from different 3D trajectories. Here, we have developed a deep learning method for super-resolved astigmatism-based 3D localization of chromosomal loci in live E. coli cells which enables a precision better than 61 nm at a signal-to-background ratio of ~4 on a heterogeneous cell background. Determining the spatial localization of chromosomal loci, we find that some loci are at the periphery of the nucleoid for large parts of the cell cycle. Analyses of individual trajectories reveal that these loci are subdiffusive both longitudinally (x) and radially (r), but that individual loci explore the full radial width on a minute time scale.