Specific connectivity optimizes learning in thalamocortical loops

Thalamocortical loops have a central role in cognition and motor control, but precisely how they contribute to these processes is unclear. Recent studies showing evidence of plasticity in thalamocortical synapses indicate a role for the thalamus in shaping cortical dynamics through learning. Since signals undergo a compression from the cortex to the thalamus, we hypothesized that the computational role of the thalamus depends critically on the structure of corticothalamic connectivity. To test this, we identified the optimal corticothalamic structure that promotes biologically plausible learning in thalamocortical synapses. We found that corticothalamic projections specialized to communicate an efference copy of the cortical output benefit motor control, while communicating the modes of highest variance is optimal for working memory tasks. We analyzed neural recordings from mice performing grasping and delayed discrimination tasks and found corticothalamic communication consistent with these predictions. These results suggest that the thalamus orchestrates cortical dynamics in a functionally precise manner through structured connectivity.

Exploring Types of Photonic Neural Networks for Imaging and Computing-A Review

Photonic neural networks (PNNs), utilizing light-based technologies, show immense potential in artificial intelligence (AI) and computing. Compared to traditional electronic neural networks, they offer faster processing speeds, lower energy usage, and improved parallelism. Leveraging light's properties for information processing could revolutionize diverse applications, including complex calculations and advanced machine learning (ML). Furthermore, these networks could address scalability and efficiency challenges in large-scale AI systems, potentially reshaping the future of computing and AI research. In this comprehensive review, we provide current, cutting-edge insights into diverse types of PNNs crafted for both imaging and computing purposes. Additionally, we delve into the intricate challenges they encounter during implementation, while also illuminating the promising perspectives they introduce to the field.

Physical Activity Detection for Diabetes Mellitus Patients Using Recurrent Neural Networks

Diabetes mellitus (DM) is a persistent metabolic disorder associated with the hormone insulin. The two main types of DM are type 1 (T1DM) and type 2 (T2DM). Physical activity plays a crucial role in the therapy of diabetes, benefiting both types of patients. The detection, recognition, and subsequent classification of physical activity based on type and intensity are integral components of DM treatment. The continuous glucose monitoring system (CGMS) signal provides the blood glucose (BG) level, and the combination of CGMS and heart rate (HR) signals are potential targets for detecting relevant physical activity from the BG variation point of view. The main objective of the present research is the developing of an artificial intelligence (AI) algorithm capable of detecting physical activity using these signals. Using multiple recurrent models, the best-achieved performance of the different classifiers is a 0.99 area under the receiver operating characteristic curve. The application of recurrent neural networks (RNNs) is shown to be a powerful and efficient solution for accurate detection and analysis of physical activity in patients with DM. This approach has great potential to improve our understanding of individual activity patterns, thus contributing to a more personalized and effective management of DM.

Recurrent neural network for pitch control of variable-speed wind turbine

Wind is one of the most widely used renewable energy sources due to its cost-effectiveness, power requirements, operation, and performance. There are many challenges in wind turbines, such as wind fluctuation, pitch control, and generator speed control. When the wind speed exceeds its rated value, the pitch angle controller limits the generator output power to its rated value. In this research work, several soft computing techniques have been implemented for pitch control of variable-speed wind turbine. The data is collected for the National Renewable Energy Laboratory offshore 5 MW baseline wind turbine. Wind speed, tip speed ratio, and power coefficient are taken as inputs, and pitch angle as output. Machine learning and artificial intelligence-based techniques such as recurrent neural networks (RNNs), adaptive neuro-fuzzy inference system (ANFIS), multilayer perceptron feed-forward neural network (MLPFFNN), and fuzzy logic controller (FLC) are implemented on MATLAB, and their results are evaluated in terms of mean square error (MSE) and root mean square error (RMSE). The controllers have been implemented in MATLAB/Simulink to schedule the wind turbine blade pitch angle and keep the output power stable at the rated value. The experimental results show that RNN provided the best results for 15 neurons in hidden layers and 1000 epochs with MSE of 3.28e-11 and RMSE of 5.54e-06, followed by MLPFFNN with MSE of 2.17e-10 and RMSE of 1.56e-05, ANFIS with MSE of 8.5e-05 and RMSE of 9.22e-03, and FLC with MSE of 6.25e-04 and RMSE of 0.025. The proposed scheme is more reliable and robust and can be easily implemented on a physical setup by using interfacing cards such as dSPACE, NI cards, and data acquisition cards.

Advanced detection of coronary artery disease via deep learning analysis of plasma cytokine data

The 2017 World Health Organization Fact Sheet highlights that coronary artery disease is the leading cause of death globally, responsible for approximately 30% of all deaths. In this context, machine learning (ML) technology is crucial in identifying coronary artery disease, thereby saving lives. ML algorithms can potentially analyze complex patterns and correlations within medical data, enabling early detection and accurate diagnosis of CAD. By leveraging ML technology, healthcare professionals can make informed decisions and implement timely interventions, ultimately leading to improved outcomes and potentially reducing the mortality rate associated with coronary artery disease. Machine learning algorithms create non-invasive, quick, accurate, and economical diagnoses. As a result, machine learning algorithms can be employed to supplement existing approaches or as a forerunner to them. This study shows how to use the CNN classifier and RNN based on the LSTM classifier in deep learning to attain targeted "risk" CAD categorization utilizing an evolving set of 450 cytokine biomarkers that could be used as suggestive solid predictive variables for treatment. The two used classifiers are based on these "45" different cytokine prediction characteristics. The best Area Under the Receiver Operating Characteristic curve (AUROC) score achieved is (0.98) for a confidence interval (CI) of 95; the classifier RNN-LSTM used "450" cytokine biomarkers had a great (AUROC) score of 0.99 with a confidence interval of 0.95 the percentage 95, the CNN model containing cytokines received the second best AUROC score (0.92). The RNN-LSTM classifier considerably beats the CNN classifier regarding AUROC scores, as evidenced by a p-value smaller than 7.48 obtained via an independent t-test. As large-scale initiatives to achieve early, rapid, reliable, inexpensive, and accessible individual identification of CAD risk gain traction, robust machine learning algorithms can now augment older methods such as angiography. Incorporating 65 new sensitive cytokine biomarkers can increase early detection even more. Investigating the novel involvement of cytokines in CAD could lead to better risk detection, disease mechanism discovery, and new therapy options.

A Hybrid Convolutional and Recurrent Neural Network for Multi-Sensor Pile Damage Detection with Time Series

Machine learning (ML) algorithms are increasingly applied to structure health monitoring (SHM) problems. However, their application to pile damage detection (PDD) is hindered by the complexity of the problem. A novel multi-sensor pile damage detection (MSPDD) method is proposed in this paper to extend the application of ML algorithms in the automatic identification of PDD. The time-series signals collected by multiple sensors during the pile integrity test are first processed by the traveling wave decomposition (TWD) theory and are then input into a hybrid one-dimensional (1D) convolutional and recurrent neural network. The hybrid neural network can achieve the automatic multi-task identification of pile damage detection based on the time series of MSPDD results. Finally, the analytical solution-based sample set is utilized to evaluate the performance of the proposed hybrid model. The outputs of the multi-task learning framework can provide a detailed description of the actual pile quality and provide strong support for the classification of pile quality as well.

MRI-Based Kinetic Heterogeneity Evaluation in the Accurate Access of Axillary Lymph Node Status in Breast Cancer Using a Hybrid CNN-RNN Model

BACKGROUND

Accurate evaluation of the axillary lymph node (ALN) status is needed for determining the treatment protocol for breast cancer (BC). The value of magnetic resonance imaging (MRI)-based tumor heterogeneity in assessing ALN metastasis in BC is unclear.

PURPOSE

To assess the value of deep learning (DL)-derived kinetic heterogeneity parameters based on BC dynamic contrast-enhanced (DCE)-MRI to infer the ALN status.

STUDY TYPE

Retrospective.

SUBJECTS

1256/539/153/115 patients in the training cohort, internal validation cohort, and external validation cohorts I and II, respectively.

FIELD STRENGTH/SEQUENCE

1.5 T/3.0 T, non-contrast T1-weighted spin-echo sequence imaging (T1WI), DCE-T1WI, and diffusion-weighted imaging.

ASSESSMENT

Clinical pathological and MRI semantic features were obtained by reviewing histopathology and MRI reports. The segmentation of the tumor lesion on the first phase of T1WI DCE-MRI images was applied to other phases after registration. A DL architecture termed convolutional recurrent neural network (ConvRNN) was developed to generate the KHimage (kinetic heterogeneity of DCE-MRI image) score that indicated the ALN status in patients with BC. The model was trained and optimized on training and internal validation cohorts, tested on two external validation cohorts. We compared ConvRNN model with other 10 models and the subgroup analyses of tumor size, magnetic field strength, and molecular subtype were also evaluated.

STATISTICAL TESTS

Chi-squared, Fisher's exact, Student's t, Mann-Whitney U tests, and receiver operating characteristics (ROC) analysis were performed. P < 0.05 was considered significant.

RESULTS

The ConvRNN model achieved area under the curve (AUC) of 0.802 in the internal validation cohort and 0.785-0.806 in the external validation cohorts. The ConvRNN model could well evaluate the ALN status of the four molecular subtypes (AUC = 0.685-0.868). The patients with larger tumor sizes (>5 cm) were more susceptible to ALN metastasis with KHimage scores of 0.527-0.827.

DATA CONCLUSION

A ConvRNN model outperformed traditional models for determining the ALN status in patients with BC.

LEVEL OF EVIDENCE

3 TECHNICAL EFFICACY: Stage 2.

Electrical micro flow cytometry with LSTM and its application in leukocyte differential

This paper developed an electrical micro flow cytometry to realize leukocyte differentials leveraging a constrictional microchannel and a deep neural network. Firstly, purified granulocytes, lymphocytes or monocytes traveled through the constrictional microchannel with a cross-sectional area marginally larger than individual cells and produced large impedance variations by blocking focused electric field lines. By optimizing key elements (e.g., normalization, learning rate, batch size and neuron number) of the recurrent neural network (RNN), electrical results of purified leukocytes were analyzed to establish a leukocyte differential system with a classification accuracy of 95.2%. Then the leukocyte mixtures were forced to travel through the same constrictional microchannel, producing mixed impedance profiles which were classified into granulocytes, lymphocytes and monocytes based on the aforementioned differential system. As to the classification results, two leukocyte mixtures from the same donor were processed, producing comparable classification results, which were 57% versus 59% of granulocytes, 37% versus 34% of lymphocytes and 6% versus 7% of monocytes. These results validated the established classification system based on the constrictional microchannel and the recurrent neural network, providing a new perspective of differentiating white blood cells by electrical flow cytometry.

A Molecular Generative Model of COVID-19 Main Protease Inhibitors Using Long Short-Term Memory-Based Recurrent Neural Network

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused a serious threat to public health and prompted researchers to find anti-coronavirus 2019 (COVID-19) compounds. In this study, the long short-term memory-based recurrent neural network was used to generate new inhibitors for the coronavirus. First, the model was trained to generate drug compounds in the form of valid simplified molecular-input line-entry system strings. Then, the structures of COVID-19 main protease inhibitors were applied to fine-tune the model. After fine-tuning, the network could generate new molecular structures as novel SARS-CoV-2 main protease inhibitors. Molecular docking exhibited that some generated compounds have the proper affinity to the active site of the protease. Molecular Dynamics simulations explored binding free energies of the compounds over simulation trajectories. In addition, in silico absorption, distribution, metabolism, and excretion studies showed that some novel compounds could be formulated as orally active agents. Based on molecular docking and molecular dynamics simulation studies, compound AADH possessed significant binding affinity and presumably inhibition against the SARS-CoV-2 main protease enzyme. Therefore, the proposed deep learning-based model was capable of generating promising anti-COVID-19 drugs.

An ensemble-based serial cascaded attention network and improved variational auto encoder for breast cancer prognosis prediction using data

Breast cancer is one of the most common types of cancer in women and it produces a huge amount of death rate in the world. Early recognition is lessening its impact. The early recognition of breast cancer could convince patients to receive surgical therapy, which will significantly improve the chance of restoration. This information is used by the machine learning technique to find links between them and appraise our forecasts of fresh occurrences. Later recognition of breast cancer can lead to death. An accurate prescient framework for breast cancer prediction is urgently needed in the current era. In order to accomplish the objective, an adaptive ensemble model is proposed for breast cancer prognosis prediction using data. At the initial stage, the raw data are fetched from benchmark datasets. It is then followed by data cleaning and preprocessing. Subsequently, the pre-processed data is fed into the Improved Variational Autoencoder (IVAE), where the deep features are extracted. Finally, the resultant features are given as input to the Ensemble-based Serial Cascaded Attention Network (ESCANet), which is built with Deep Temporal Convolution Network (DTCN), Bi-directional Long Short-Term Memory (BiLSTM), and Recurrent Neural Network (RNN). The effectiveness of the model is validated and compared with conventional methodologies. Therefore, the results elucidate that the proposed methodology achieves extensive results; thus, it increases the system's efficiency.

LSTM-Based Virtual Load Sensor for Heavy-Duty Vehicles

In this paper, a special recurrent neural network (RNN) called Long Short-Term Memory (LSTM) is used to design a virtual load sensor that estimates the mass of heavy vehicles. The estimation algorithm consists of a two-layer LSTM network. The network estimates vehicle mass based on vehicle speed, longitudinal acceleration, engine speed, engine torque, and accelerator pedal position. The network is trained and tested with a data set collected in a high-fidelity simulation environment called Truckmaker. The training data are generated in acceleration maneuvers across a range of speeds, while the test data are obtained by simulating the vehicle in the Worldwide harmonized Light vehicles Test Cycle (WLTC). Preliminary results show that, with the proposed approach, heavy-vehicle mass can be estimated as accurately as commercial load sensors across a range of load mass as wide as four tons.

Biosensor-Based Multimodal Deep Human Locomotion Decoding via Internet of Healthcare Things

Multiple Internet of Healthcare Things (IoHT)-based devices have been utilized as sensing methodologies for human locomotion decoding to aid in applications related to e-healthcare. Different measurement conditions affect the daily routine monitoring, including the sensor type, wearing style, data retrieval method, and processing model. Currently, several models are present in this domain that include a variety of techniques for pre-processing, descriptor extraction, and reduction, along with the classification of data captured from multiple sensors. However, such models consisting of multiple subject-based data using different techniques may degrade the accuracy rate of locomotion decoding. Therefore, this study proposes a deep neural network model that not only applies the state-of-the-art Quaternion-based filtration technique for motion and ambient data along with background subtraction and skeleton modeling for video-based data, but also learns important descriptors from novel graph-based representations and Gaussian Markov random-field mechanisms. Due to the non-linear nature of data, these descriptors are further utilized to extract the codebook via the Gaussian mixture regression model. Furthermore, the codebook is provided to the recurrent neural network to classify the activities for the locomotion-decoding system. We show the validity of the proposed model across two publicly available data sampling strategies, namely, the HWU-USP and LARa datasets. The proposed model is significantly improved over previous systems, as it achieved 82.22% and 82.50% for the HWU-USP and LARa datasets, respectively. The proposed IoHT-based locomotion-decoding model is useful for unobtrusive human activity recognition over extended periods in e-healthcare facilities.

Data-driven crash prediction by injury severity using a recurrent neural network model based on Keras framework

With the development of big data technology and the improvement of deep learning technology, data-driven and machine learning application have been widely employed. By adopting the data-driven machine learning method, with the help of clustering processing of data sets, a recurrent neural network (RNN) model based on Keras framework is proposed to predict the injury severity in urban areas. First, with crash data from 2014 to 2017 in Nevada, OPTICS clustering algorithm is employed to extract the crash injury in Las Vegas. Next, by virtue of Keras' high efficiency and strong scalability, the parameters of loss function, activation function and optimizer of the deep learning model are determined to realize the training of the model and the visualization of the training results, and the RNN model is constructed. Finally, on the basis of training and testing data, the model can predict the injury severity with high accuracy and high training speed. The results provide an alternative and some potential insights on the injury severity prediction.

Recurrent Neural Network Methods for Extracting Dynamic Balance Variables during Gait from a Single Inertial Measurement Unit

Monitoring dynamic balance during gait is critical for fall prevention in the elderly. The current study aimed to develop recurrent neural network models for extracting balance variables from a single inertial measurement unit (IMU) placed on the sacrum during walking. Thirteen healthy young and thirteen healthy older adults wore the IMU during walking and the ground truth of the inclination angles (IA) of the center of pressure to the center of mass vector and their rates of changes (RCIA) were measured simultaneously. The IA, RCIA, and IMU data were used to train four models (uni-LSTM, bi-LSTM, uni-GRU, and bi-GRU), with 10% of the data reserved to evaluate the model errors in terms of the root-mean-squared errors (RMSEs) and percentage relative RMSEs (rRMSEs). Independent t-tests were used for between-group comparisons. The sensitivity, specificity, and Pearson's r for the effect sizes between the model-predicted data and experimental ground truth were also obtained. The bi-GRU with the weighted MSE model was found to have the highest prediction accuracy, computational efficiency, and the best ability in identifying statistical between-group differences when compared with the ground truth, which would be the best choice for the prolonged real-life monitoring of gait balance for fall risk management in the elderly.

Leg-body coordination strategies for obstacle avoidance and narrow space navigation of multi-segmented, legged robots

Introduction

Millipedes can avoid obstacle while navigating complex environments with their multi-segmented body. Biological evidence indicates that when the millipede navigates around an obstacle, it first bends the anterior segments of its corresponding anterior segment of its body, and then gradually propagates this body bending mechanism from anterior to posterior segments. Simultaneously, the stride length between pairs of legs inside the bending curve decreases to coordinate the leg motions with the bending mechanism of the body segments. In robotics, coordination between multiple legs and body segments during turning for navigating in complex environments, e.g., narrow spaces, has not been fully realized in multi-segmented, multi-legged robots with more than six legs.

Method

To generate the efficient obstacle avoidance turning behavior in a multi-segmented, multi-legged (millipede-like) robot, this study explored three possible strategies of leg and body coordination during turning: including the local leg and body coordination at the segment level in a manner similar to millipedes, global leg amplitude change in response to different turning directions (like insects), and the phase reversal of legs inside of turning curve during obstacle avoidance (typical engineering approach).

Results

Using sensory inputs obtained from the antennae located at the robot head and recurrent neural control, different turning strategies were generated, with gradual body bending propagation from the anterior to posterior body segments.

Discussion

We discovered differences in the performance of each turning strategy, which could guide the future control development of multi-segmented, legged robots.

Prediction for blood lactate during exercise using an artificial intelligence-Enabled electrocardiogram: a feasibility study

Introduction: The acquisition of blood lactate concentration (BLC) during exercise is beneficial for endurance training, yet a convenient method to measure it remains unavailable. BLC and electrocardiogram (ECG) both exhibit variations with changes in exercise intensity and duration. In this study, we hypothesized that BLC during exercise can be predicted using ECG data. Methods: Thirty-one healthy participants underwent four cardiopulmonary exercise tests, including one incremental test and three constant work rate (CWR) tests at low, moderate, and high intensity. Venous blood samples were obtained immediately after each CWR test to measure BLC. A mathematical model was constructed using 31 trios of CWR tests, which utilized a residual network combined with long short-term memory to analyze every beat of lead II ECG waveform as 2D images. An artificial neural network was used to analyze variables such as the RR interval, age, sex, and body mass index. Results: The standard deviation of the fitting error was 0.12 mmol/L for low and moderate intensities, and 0.19 mmol/L for high intensity. Weighting analysis demonstrated that ECG data, including every beat of ECG waveform and RR interval, contribute predominantly. Conclusion: By employing 2D convolution and artificial neural network-based methods, BLC during exercise can be accurately estimated non-invasively using ECG data, which has potential applications in exercise training.

Ensemble Approach to Combining Episode Prediction Models Using Sequential Circadian Rhythm Sensor Data from Mental Health Patients

Managing mood disorders poses challenges in counseling and drug treatment, owing to limitations. Counseling is the most effective during hospital visits, and the side effects of drugs can be burdensome. Patient empowerment is crucial for understanding and managing these triggers. The daily monitoring of mental health and the utilization of episode prediction tools can enable self-management and provide doctors with insights into worsening lifestyle patterns. In this study, we test and validate whether the prediction of future depressive episodes in individuals with depression can be achieved by using lifelog sequence data collected from digital device sensors. Diverse models such as random forest, hidden Markov model, and recurrent neural network were used to analyze the time-series data and make predictions about the occurrence of depressive episodes in the near future. The models were then combined into a hybrid model. The prediction accuracy of the hybrid model was 0.78; especially in the prediction of rare episode events, the F1-score performance was approximately 1.88 times higher than that of the dummy model. We explored factors such as data sequence size, train-to-test data ratio, and class-labeling time slots that can affect the model performance to determine the combinations of parameters that optimize the model performance. Our findings are especially valuable because they are experimental results derived from large-scale participant data analyzed over a long period of time.

Protein-Protein Interaction Network Extraction Using Text Mining Methods Adds Insight into Autism Spectrum Disorder

Text mining methods are being developed to assimilate the volume of biomedical textual materials that are continually expanding. Understanding protein-protein interaction (PPI) deficits would assist in explaining the genesis of diseases. In this study, we designed an automated system to extract PPIs from the biomedical literature that uses a deep learning sentence classification model, a pretrained word embedding, and a BiLSTM recurrent neural network with additional layers, a conditional random field (CRF) named entity recognition (NER) model, and shortest-dependency path (SDP) model using the SpaCy library in Python. The automated system ensures that it targets sentences that contain PPIs and not just these proteins mentioned in the framework of disease discovery or other context. Our first model achieved 13% greater precision on the Aimed/BioInfr benchmark corpus than the previous state-of-the-art BiLSTM neural network models. The NER model presented in this study achieved 98% precision on the Aimed/BioInfr corpus over previous models. In order to facilitate the production of an accurate representation of the PPI network, the processes were developed to systematically map the protein interactions in the texts. Overall, evaluating our system through the use of 6027 abstracts pertaining to seven proteins associated with Autism Spectrum Disorder completed the manually curated PPI network for these proteins. When it comes to complicated diseases, these networks would assist in understanding how PPI deficits contribute to disease development while also emphasizing the influence of interactions on protein function and biological processes.

CARDIAN: a novel computational approach for real-time end-diastolic frame detection in intravascular ultrasound using bidirectional attention networks

Introduction

Changes in coronary artery luminal dimensions during the cardiac cycle can impact the accurate quantification of volumetric analyses in intravascular ultrasound (IVUS) image studies. Accurate ED-frame detection is pivotal for guiding interventional decisions, optimizing therapeutic interventions, and ensuring standardized volumetric analysis in research studies. Images acquired at different phases of the cardiac cycle may also lead to inaccurate quantification of atheroma volume due to the longitudinal motion of the catheter in relation to the vessel. As IVUS images are acquired throughout the cardiac cycle, end-diastolic frames are typically identified retrospectively by human analysts to minimize motion artefacts and enable more accurate and reproducible volumetric analysis.

Methods

In this paper, a novel neural network-based approach for accurate end-diastolic frame detection in IVUS sequences is proposed, trained using electrocardiogram (ECG) signals acquired synchronously during IVUS acquisition. The framework integrates dedicated motion encoders and a bidirectional attention recurrent network (BARNet) with a temporal difference encoder to extract frame-by-frame motion features corresponding to the phases of the cardiac cycle. In addition, a spatiotemporal rotation encoder is included to capture the IVUS catheter's rotational movement with respect to the coronary artery.

Results

With a prediction tolerance range of 66.7 ms, the proposed approach was able to find 71.9%, 67.8%, and 69.9% of end-diastolic frames in the left anterior descending, left circumflex and right coronary arteries, respectively, when tested against ECG estimations. When the result was compared with two expert analysts' estimation, the approach achieved a superior performance.

Discussion

These findings indicate that the developed methodology is accurate and fully reproducible and therefore it should be preferred over experts for end-diastolic frame detection in IVUS sequences.

A Review of Recurrent Neural Network-Based Methods in Computational Physiology

Artificial intelligence and machine learning techniques have progressed dramatically and become powerful tools required to solve complicated tasks, such as computer vision, speech recognition, and natural language processing. Since these techniques have provided promising and evident results in these fields, they emerged as valuable methods for applications in human physiology and healthcare. General physiological recordings are time-related expressions of bodily processes associated with health or morbidity. Sequence classification, anomaly detection, decision making, and future status prediction drive the learning algorithms to focus on the temporal pattern and model the nonstationary dynamics of the human body. These practical requirements give birth to the use of recurrent neural networks (RNNs), which offer a tractable solution in dealing with physiological time series and provide a way to understand complex time variations and dependencies. The primary objective of this article is to provide an overview of current applications of RNNs in the area of human physiology for automated prediction and diagnosis within different fields. Finally, we highlight some pathways of future RNN developments for human physiology.

A Study on Survival Analysis Methods Using Neural Network to Prevent Cancers

Background: Cancer is one of the main global health threats. Early personalized prediction of cancer incidence is crucial for the population at risk. This study introduces a novel cancer prediction model based on modern recurrent survival deep learning algorithms. Methods: The study includes 160,407 participants from the blood-based cohort of the Korea Cancer Prevention Research-II Biobank, which has been ongoing since 2004. Data linkages were designed to ensure anonymity, and data collection was carried out through nationwide medical examinations. Predictive performance on ten cancer sites, evaluated using the concordance index (c-index), was compared among nDeep and its multitask variation, Cox proportional hazard (PH) regression, DeepSurv, and DeepHit. Results: Our models consistently achieved a c-index of over 0.8 for all ten cancers, with a peak of 0.8922 for lung cancer. They outperformed Cox PH regression and other survival deep neural networks. Conclusion: This study presents a survival deep learning model that demonstrates the highest predictive performance on censored health dataset, to the best of our knowledge. In the future, we plan to investigate the causal relationship between explanatory variables and cancer to reduce cancer incidence and mortality.

Artificial Intelligence-Based Secured Power Grid Protocol for Smart City

Due to the modern power system's rapid development, more scattered smart grid components are securely linked into the power system by encircling a wide electrical power network with the underpinning communication system. By enabling a wide range of applications, such as distributed energy management, system state forecasting, and cyberattack security, these components generate vast amounts of data that automate and improve the efficiency of the smart grid. Due to traditional computer technologies' inability to handle the massive amount of data that smart grid systems generate, AI-based alternatives have received a lot of interest. Long Short-Term Memory (LSTM) and recurrent Neural Networks (RNN) will be specifically developed in this study to address this issue by incorporating the adaptively time-developing energy system's attributes to enhance the model of the dynamic properties of contemporary Smart Grid (SG) that are impacted by Revised Encoding Scheme (RES) or system reconfiguration to differentiate LSTM changes & real-time threats. More specifically, we provide a federated instructional strategy for consumer sharing of power data to Power Grid (PG) that is supported by edge clouds, protects consumer privacy, and is communication-efficient. They then design two optimization problems for Energy Data Owners (EDO) and energy service operations, as well as a local information assessment method in Federated Learning (FL) by taking non-independent and identically distributed (IID) effects into consideration. The test results revealed that LSTM had a longer training duration, four hidden levels, and higher training loss than other models. The provided method works incredibly well in several situations to identify FDIA. The suggested approach may successfully induce EDOs to employ high-quality local models, increase the payout of the ESP, and decrease task latencies, according to extensive simulations, which are the last points. According to the verification results, every assault sample could be effectively recognized utilizing the current detection methods and the LSTM RNN-based structure created by Smart.

Use of recurrent neural networks considering maintenance to predict urban road performance in Beijing, China

A correct understanding of the pavement performance change law forms the premise of the scientific formulation of maintenance decisions. This paper aims to develop a predictive model taking into account the costs of different types of maintenance works that reflects the continuous true usage performance of the pavement. The model proposed in this study was trained on a dataset containing five-year maintenance work data on urban roads in Beijing with pavement performance indicators for the corresponding years. The same roads were matched and combined to obtain a set of sequences of pavement performance changes with the features of the current year; with the recurrent-neural-network-based long short-term memory (LSTM) network and gate recurrent unit (GRU) network, the prediction accuracy of highway pavement performance on the test set was significantly increased. The prediction result indicates that the generalization ability of the improved recurrent neural network model is satisfactory, with the R² achieving 0.936, and of the two models the GRU model is more efficient, with an accuracy that reaches almost the same level as LSTM but with the training convergence time reduced to 25 s. This study demonstrates that data generated by the work of maintenance units can be used effectively in the prediction of pavement performance. This article is part of the theme issue 'Artificial intelligence in failure analysis of transportation infrastructure and materials'.

A deep learning-based dynamic model for predicting acute kidney injury risk severity in postoperative patients

BACKGROUND

Acute kidney injury is a common postoperative complication affecting between 10% and 30% of surgical patients. Acute kidney injury is associated with increased resource usage and chronic kidney disease development, with more severe acute kidney injury suggesting more aggressive deterioration in clinical outcomes and mortality.

METHODS

We considered 42,906 surgical patients admitted to University of Florida Health (n = 51,806) between 2014 and 2021. Acute kidney injury stages were determined using the Kidney Disease Improving Global Outcomes serum creatinine criteria. We developed a recurrent neural network-based model to continuously predict acute kidney injury risk and state in the following 24 hours and compared it with logistic regression, random forest, and multi-layer perceptron models. We used medications, laboratory and vital measurements, and derived features from past one-year records as inputs. We analyzed the proposed model with integrated gradients for enhanced explainability.

RESULTS

Postoperative acute kidney injury at any stage developed in 20% (10,664) of the cohort. The recurrent neural network model was more accurate in predicting nearly all categories of next-day acute kidney injury stages (including the no acute kidney injury group). The area under the receiver operating curve and 95% confidence intervals for recurrent neural network and logistic regression models were for no acute kidney injury (0.98 [0.98-0.98] vs 0.93 [0.93-0.93]), stage 1 (0.95 [0.95-0.95] vs. 0.81 [0.80-0.82]), stage 2/3 (0.99 [0.99-0.99] vs 0.96 [0.96-0.97]), and stage 3 with renal replacement therapy (1.0 [1.0-1.0] vs 1.0 [1.0-1.0].

CONCLUSION

The proposed model demonstrates that temporal processing of patient information can lead to more granular and dynamic modeling of acute kidney injury status and result in more continuous and accurate acute kidney injury prediction. We showcase the integrated gradients framework's utility as a mechanism for enhancing model explainability, potentially facilitating clinical trust for future implementation.

Effects of Data Augmentation on the Nine-Axis IMU-Based Orientation Estimation Accuracy of a Recurrent Neural Network

The nine-axis inertial and measurement unit (IMU)-based three-dimensional (3D) orientation estimation is a fundamental part of inertial motion capture. Recently, owing to the successful utilization of deep learning in various applications, orientation estimation neural networks (NNs) trained on large datasets, including nine-axis IMU signals and reference orientation data, have been developed. During the training process, the limited amount of training data is a critical issue in the development of powerful networks. Data augmentation, which increases the amount of training data, is a key approach for addressing the data shortage problem and thus for improving the estimation performance. However, to the best of our knowledge, no studies have been conducted to analyze the effects of data augmentation techniques on estimation performance in orientation estimation networks using IMU sensors. This paper selects three data augmentation techniques for IMU-based orientation estimation NNs, i.e., augmentation by virtual rotation, bias addition, and noise addition (which are hereafter referred to as rotation, bias, and noise, respectively). Then, this paper analyzes the effects of these augmentation techniques on estimation accuracy in recurrent neural networks, for a total of seven combinations (i.e., rotation only, bias only, noise only, rotation and bias, rotation and noise, and rotation and bias and noise). The evaluation results show that, among a total of seven augmentation cases, four cases including 'rotation' (i.e., rotation only, rotation and bias, rotation and noise, and rotation and bias and noise) occupy the top four. Therefore, it may be concluded that the augmentation effect of rotation is overwhelming compared to those of bias and noise. By applying rotation augmentation, the performance of the NN can be significantly improved. The analysis of the effect of the data augmentation techniques presented in this paper may provide insights for developing robust IMU-based orientation estimation networks.

Localized estimation of electromagnetic sources underlying event-related fields using recurrent neural networks

Objective. To use a recurrent neural network (RNN) to reconstruct neural activity responsible for generating noninvasively measured electromagnetic signals.Approach. Output weights of an RNN were fixed as the lead field matrix from volumetric source space computed using the boundary element method with co-registered structural magnetic resonance images and magnetoencephalography (MEG). Initially, the network was trained to minimise mean-squared-error loss between its outputs and MEG signals, causing activations in the penultimate layer to converge towards putative neural source activations. Subsequently, L1 regularisation was applied to the final hidden layer, and the model was fine-tuned, causing it to favour more focused activations. Estimated source signals were then obtained from the outputs of the last hidden layer. We developed and validated this approach with simulations before applying it to real MEG data, comparing performance with beamformers, minimum-norm estimate, and mixed-norm estimate source reconstruction methods.Main results. The proposed RNN method had higher output signal-to-noise ratios and comparable correlation and error between estimated and simulated sources. Reconstructed MEG signals were also equal or superior to the other methods regarding their similarity to ground-truth. When applied to MEG data recorded during an auditory roving oddball experiment, source signals estimated with the RNN were generally biophysically plausible and consistent with expectations from the literature.Significance. This work builds on recent developments of RNNs for modelling event-related neural responses by incorporating biophysical constraints from the forward model, thus taking a significant step towards greater biological realism and introducing the possibility of exploring how input manipulations may influence localised neural activity.

Discovering anomalies in big data: a review focused on the application of metaheuristics and machine learning techniques

With the increase in available data from computer systems and their security threats, interest in anomaly detection has increased as well in recent years. The need to diagnose faults and cyberattacks has also focused scientific research on the automated classification of outliers in big data, as manual labeling is difficult in practice due to their huge volumes. The results obtained from data analysis can be used to generate alarms that anticipate anomalies and thus prevent system failures and attacks. Therefore, anomaly detection has the purpose of reducing maintenance costs as well as making decisions based on reports. During the last decade, the approaches proposed in the literature to classify unknown anomalies in log analysis, process analysis, and time series have been mainly based on machine learning and deep learning techniques. In this study, we provide an overview of current state-of-the-art methodologies, highlighting their advantages and disadvantages and the new challenges. In particular, we will see that there is no absolute best method, i.e., for any given dataset a different method may achieve the best result. Finally, we describe how the use of metaheuristics within machine learning algorithms makes it possible to have more robust and efficient tools.

The Use of Respiratory Effort Improves an ECG-Based Deep Learning Algorithm to Assess Sleep-Disordered Breathing

BACKGROUND

Sleep apnea is a prevalent sleep-disordered breathing (SDB) condition that affects a large population worldwide. Research has demonstrated the potential of using electrocardiographic (ECG) signals (heart rate and ECG-derived respiration, EDR) to detect SDB. However, EDR may be a suboptimal replacement for respiration signals.

METHODS

We evaluated a previously described ECG-based deep learning algorithm in an independent dataset including 198 patients and compared performance for SDB event detection using thoracic respiratory effort versus EDR. We also evaluated the algorithm in terms of apnea-hypopnea index (AHI) estimation performance, and SDB severity classification based on the estimated AHI.

RESULTS

Using respiratory effort instead of EDR, we achieved an improved performance in SDB event detection (F1 score = 0.708), AHI estimation (Spearman's correlation = 0.922), and SDB severity classification (Cohen's kappa of 0.62 was obtained based on AHI).

CONCLUSION

Respiratory effort is superior to EDR to assess SDB. Using respiratory effort and ECG, the previously described algorithm achieves good performance in a new dataset from an independent laboratory confirming its adequacy for this task.

Insilico generation of novel ligands for the inhibition of SARS-CoV-2 main protease (3CLpro) using deep learning

The recent emergence of novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) causing the coronavirus disease (COVID-19) has become a global public health crisis, and a crucial need exists for rapid identification and development of novel therapeutic interventions. In this study, a recurrent neural network (RNN) is trained and optimized to produce novel ligands that could serve as potential inhibitors to the SARS-CoV-2 viral protease: 3 chymotrypsin-like protease (3CL^pro). Structure-based virtual screening was performed through molecular docking, ADMET profiling, and predictions of various molecular properties were done to evaluate the toxicity and drug-likeness of the generated novel ligands. The properties of the generated ligands were also compared with current drugs under various phases of clinical trials to assess the efficacy of the novel ligands. Twenty novel ligands were selected that exhibited good drug-likeness properties, with most ligands conforming to Lipinski's rule of 5, high binding affinity (highest binding affinity: -9.4 kcal/mol), and promising ADMET profile. Additionally, the generated ligands complexed with 3CL^pro were found to be stable based on the results of molecular dynamics simulation studies conducted over a 100 ns period. Overall, the findings offer a promising avenue for the rapid identification and development of effective therapeutic interventions to treat COVID-19.

Fog-Based Smart Cardiovascular Disease Prediction System Powered by Modified Gated Recurrent Unit

The ongoing fast-paced technology trend has brought forth ceaseless transformation. In this regard, cloud computing has long proven to be the paramount deliverer of services such as computing power, software, networking, storage, and databases on a pay-per-use basis. The cloud is a big proponent of the internet of things (IoT), furnishing the computation and storage requisite to address internet-of-things applications. With the proliferating IoT devices triggering a continual data upsurge, the cloud-IoT interaction encounters latency, bandwidth, and connectivity restraints. The inclusion of the decentralized and distributed fog computing layer amidst the cloud and IoT layer extends the cloud's processing, storage, and networking services close to end users. This hierarchical edge-fog-cloud model distributes computation and intelligence, yielding optimal solutions while tackling constraints like massive data volume, latency, delay, and security vulnerability. The healthcare domain, warranting time-critical functionalities, can reap benefits from the cloud-fog-IoT interplay. This research paper propounded a fog-assisted smart healthcare system to diagnose heart or cardiovascular disease. It combined a fuzzy inference system (FIS) with the recurrent neural network model's variant of the gated recurrent unit (GRU) for pre-processing and predictive analytics tasks. The proposed system showcases substantially improved performance results, with classification accuracy at 99.125%. With major processing of healthcare data analytics happening at the fog layer, it is observed that the proposed work reveals optimized results concerning delays in terms of latency, response time, and jitter, compared to the cloud. Deep learning models are adept at handling sophisticated tasks, particularly predictive analytics. Time-critical healthcare applications reap benefits from deep learning's exclusive potential to furnish near-perfect results, coupled with the merits of the decentralized fog model, as revealed by the experimental results.

Recurrent Network Solutions for Human Posture Recognition Based on Kinect Skeletal Data

Ambient Assisted Living (AAL) systems are designed to provide unobtrusive and user-friendly support in daily life and can be used for monitoring frail people based on various types of sensors, including wearables and cameras. Although cameras can be perceived as intrusive in terms of privacy, low-cost RGB-D devices (i.e., Kinect V2) that extract skeletal data can partially overcome these limits. In addition, deep learning-based algorithms, such as Recurrent Neural Networks (RNNs), can be trained on skeletal tracking data to automatically identify different human postures in the AAL domain. In this study, we investigate the performance of two RNN models (2BLSTM and 3BGRU) in identifying daily living postures and potentially dangerous situations in a home monitoring system, based on 3D skeletal data acquired with Kinect V2. We tested the RNN models with two different feature sets: one consisting of eight human-crafted kinematic features selected by a genetic algorithm, and another consisting of 52 ego-centric 3D coordinates of each considered skeleton joint, plus the subject's distance from the Kinect V2. To improve the generalization ability of the 3BGRU model, we also applied a data augmentation method to balance the training dataset. With this last solution we reached an accuracy of 88%, the best we achieved so far.

Software Usability Testing Using EEG-Based Emotion Detection and Deep Learning

It is becoming increasingly attractive to detect human emotions using electroencephalography (EEG) brain signals. EEG is a reliable and cost-effective technology used to measure brain activities. This paper proposes an original framework for usability testing based on emotion detection using EEG signals, which can significantly affect software production and user satisfaction. This approach can provide an in-depth understanding of user satisfaction accurately and precisely, making it a valuable tool in software development. The proposed framework includes a recurrent neural network algorithm as a classifier, a feature extraction algorithm based on event-related desynchronization and event-related synchronization analysis, and a new method for selecting EEG sources adaptively for emotion recognition. The framework results are promising, achieving 92.13%, 92.67%, and 92.24% for the valence-arousal-dominance dimensions, respectively.

Unsupervised Anomaly Detection for Cars CAN Sensors Time Series Using Small Recurrent and Convolutional Neural Networks

Predictive maintenance in the car industry is an active field of research for machine learning and anomaly detection. The capability of cars to produce time series data from sensors is growing as the car industry is heading towards more connected and electric vehicles. Unsupervised anomaly detectors are therefore very adapted to process those complex multidimensional time series and highlight abnormal behaviors. We propose to use recurrent and convolutional neural networks based on unsupervised anomaly detectors with simple architectures on real, multidimensional time series generated by the car sensors and extracted from the Controller Area Network bus (CAN). Our method is then evaluated through known specific anomalies. As the computational costs of Machine Learning algorithms are a rising issue regarding embedded scenarios such as car anomaly detection, we also focus on creating anomaly detectors that are as small as possible. Using a state-of-the-art methodology incorporating a time series predictor and a prediction-error-based anomaly detector, we show that we can obtain roughly the same anomaly detection performance with smaller predictors, reducing parameters and calculations by up to 23% and 60%, respectively. Finally, we introduce a method to correlate variables with specific anomalies by using anomaly detector results and labels.

An Enhanced Ensemble Deep Neural Network Approach for Elderly Fall Detection System Based on Wearable Sensors

Fatal injuries and hospitalizations caused by accidental falls are significant problems among the elderly. Detecting falls in real-time is challenging, as many falls occur in a short period. Developing an automated monitoring system that can predict falls before they happen, provide safeguards during the fall, and issue remote notifications after the fall is essential to improving the level of care for the elderly. This study proposed a concept for a wearable monitoring framework that aims to anticipate falls during their beginning and descent, activating a safety mechanism to minimize fall-related injuries and issuing a remote notification after the body impacts the ground. However, the demonstration of this concept in the study involved the offline analysis of an ensemble deep neural network architecture based on a Convolutional Neural Network (CNN) and a Recurrent Neural Network (RNN) and existing data. It is important to note that this study did not involve the implementation of hardware or other elements beyond the developed algorithm. The proposed approach utilized CNN for robust feature extraction from accelerometer and gyroscope data and RNN to model the temporal dynamics of the falling process. A distinct class-based ensemble architecture was developed, where each ensemble model identified a specific class. The proposed approach was evaluated on the annotated SisFall dataset and achieved a mean accuracy of 95%, 96%, and 98% for Non-Fall, Pre-Fall, and Fall detection events, respectively, outperforming state-of-the-art fall detection methods. The overall evaluation demonstrated the effectiveness of the developed deep learning architecture. This wearable monitoring system will prevent injuries and improve the quality of life of elderly individuals.

The influence of non-stationarity of spike signals on decoding performance in intracortical brain-computer interface: a simulation study

Introduction

Intracortical Brain-Computer Interfaces (iBCI) establish a new pathway to restore motor functions in individuals with paralysis by interfacing directly with the brain to translate movement intention into action. However, the development of iBCI applications is hindered by the non-stationarity of neural signals induced by the recording degradation and neuronal property variance. Many iBCI decoders were developed to overcome this non-stationarity, but its effect on decoding performance remains largely unknown, posing a critical challenge for the practical application of iBCI.

Methods

To improve our understanding on the effect of non-stationarity, we conducted a 2D-cursor simulation study to examine the influence of various types of non-stationarities. Concentrating on spike signal changes in chronic intracortical recording, we used the following three metrics to simulate the non-stationarity: mean firing rate (MFR), number of isolated units (NIU), and neural preferred directions (PDs). MFR and NIU were decreased to simulate the recording degradation while PDs were changed to simulate the neuronal property variance. Performance evaluation based on simulation data was then conducted on three decoders and two different training schemes. Optimal Linear Estimation (OLE), Kalman Filter (KF), and Recurrent Neural Network (RNN) were implemented as decoders and trained using static and retrained schemes.

Results

In our evaluation, RNN decoder and retrained scheme showed consistent better performance under small recording degradation. However, the serious signal degradation would cause significant performance to drop eventually. On the other hand, RNN performs significantly better than the other two decoders in decoding simulated non-stationary spike signals, and the retrained scheme maintains the decoders' high performance when changes are limited to PDs.

Discussion

Our simulation work demonstrates the effects of neural signal non-stationarity on decoding performance and serves as a reference for selecting decoders and training schemes in chronic iBCI. Our result suggests that comparing to KF and OLE, RNN has better or equivalent performance using both training schemes. Performance of decoders under static scheme is influenced by recording degradation and neuronal property variation while decoders under retrained scheme are only influenced by the former one.

Evaluation of Early Concrete Damage Caused by Chloride-Induced Steel Corrosion Using a Deep Learning Approach Based on RNN for Ultrasonic Pulse Waves

The objective of this study is to explore the feasibility of using ultrasonic pulse wave measurements as an early detection method for corrosion-induced concrete damages. A series of experiments are conducted using concrete cube specimens, at a size of 200 mm, with a reinforcing steel bar (rebar) embedded in the center. The main variables include the water-to-cement ratio of the concrete (0.4, 0.5, and 0.6), the diameter of the rebar (10 mm, 13 mm, 19 mm, and 22 mm), and the corrosion level (ranging from 0% to 20% depending on rebar diameter). The impressed current technique is used to accelerate corrosion of rebars in concrete immersed in a 3% NaCl solution. Ultrasonic pulse waves are collected from the concrete specimens using a pair of 50 kHz P-wave transducers in the through-transmission configuration before and after the accelerated corrosion test. Deep learning techniques, specifically three recurrent neural network (RNN) models (long short-term memory, gated recurrent unit, and bidirectional long short-term memory), are utilized to develop a classification model for early detection of concrete damage due to rebar corrosion. The performance of the RNN models is compared to conventional ultrasonic testing parameters, namely ultrasonic pulse velocity and signal consistency. The results demonstrate that the RNN method outperforms the other two methods. Among the RNN methods, the bidirectional long short-term memory RNN model had the best performance, achieving an accuracy of 74% and a Cohen's kappa coefficient of 0.48. This study establishes the potentiality of utilizing deep learning of ultrasonic pulse waves with RNN models for early detection of concrete damage associated with steel corrosion.

Evaluation of Federated Learning in Phishing Email Detection

The use of artificial intelligence (AI) to detect phishing emails is primarily dependent on large-scale centralized datasets, which has opened it up to a myriad of privacy, trust, and legal issues. Moreover, organizations have been loath to share emails, given the risk of leaking commercially sensitive information. Consequently, it has been difficult to obtain sufficient emails to train a global AI model efficiently. Accordingly, privacy-preserving distributed and collaborative machine learning, particularly federated learning (FL), is a desideratum. As it is already prevalent in the healthcare sector, questions remain regarding the effectiveness and efficacy of FL-based phishing detection within the context of multi-organization collaborations. To the best of our knowledge, the work herein was the first to investigate the use of FL in phishing email detection. This study focused on building upon a deep neural network model, particularly recurrent convolutional neural network (RNN) and bidirectional encoder representations from transformers (BERT), for phishing email detection. We analyzed the FL-entangled learning performance in various settings, including (i) a balanced and asymmetrical data distribution among organizations and (ii) scalability. Our results corroborated the comparable performance statistics of FL in phishing email detection to centralized learning for balanced datasets and low organizational counts. Moreover, we observed a variation in performance when increasing the organizational counts. For a fixed total email dataset, the global RNN-based model had a 1.8% accuracy decrease when the organizational counts were increased from 2 to 10. In contrast, BERT accuracy increased by 0.6% when increasing organizational counts from 2 to 5. However, if we increased the overall email dataset by introducing new organizations in the FL framework, the organizational level performance improved by achieving a faster convergence speed. In addition, FL suffered in its overall global model performance due to highly unstable outputs if the email dataset distribution was highly asymmetric.

Machine Learning for Optical Motion Capture-Driven Musculoskeletal Modelling from Inertial Motion Capture Data

Marker-based Optical Motion Capture (OMC) systems and associated musculoskeletal (MSK) modelling predictions offer non-invasively obtainable insights into muscle and joint loading at an in vivo level, aiding clinical decision-making. However, an OMC system is lab-based, expensive, and requires a line of sight. Inertial Motion Capture (IMC) techniques are widely-used alternatives, which are portable, user-friendly, and relatively low-cost, although with lesser accuracy. Irrespective of the choice of motion capture technique, one typically uses an MSK model to obtain the kinematic and kinetic outputs, which is a computationally expensive tool increasingly well approximated by machine learning (ML) methods. Here, an ML approach is presented that maps experimentally recorded IMC input data to the human upper-extremity MSK model outputs computed from ('gold standard') OMC input data. Essentially, this proof-of-concept study aims to predict higher-quality MSK outputs from the much easier-to-obtain IMC data. We use OMC and IMC data simultaneously collected for the same subjects to train different ML architectures that predict OMC-driven MSK outputs from IMC measurements. In particular, we employed various neural network (NN) architectures, such as Feed-Forward Neural Networks (FFNNs) and Recurrent Neural Networks (RNNs) (vanilla, Long Short-Term Memory, and Gated Recurrent Unit) and a comprehensive search for the best-fit model in the hyperparameters space in both subject-exposed (SE) as well as subject-naive (SN) settings. We observed a comparable performance for both FFNN and RNN models, which have a high degree of agreement (ravg,SE,FFNN=0.90±0.19, ravg,SE,RNN=0.89±0.17, ravg,SN,FFNN=0.84±0.23, and ravg,SN,RNN=0.78±0.23) with the desired OMC-driven MSK estimates for held-out test data. The findings demonstrate that mapping IMC inputs to OMC-driven MSK outputs using ML models could be instrumental in transitioning MSK modelling from 'lab to field'.

A novel transfer learning framework for sorghum biomass prediction using UAV-based remote sensing data and genetic markers

Yield for biofuel crops is measured in terms of biomass, so measurements throughout the growing season are crucial in breeding programs, yet traditionally time- and labor-consuming since they involve destructive sampling. Modern remote sensing platforms, such as unmanned aerial vehicles (UAVs), can carry multiple sensors and collect numerous phenotypic traits with efficient, non-invasive field surveys. However, modeling the complex relationships between the observed phenotypic traits and biomass remains a challenging task, as the ground reference data are very limited for each genotype in the breeding experiment. In this study, a Long Short-Term Memory (LSTM) based Recurrent Neural Network (RNN) model is proposed for sorghum biomass prediction. The architecture is designed to exploit the time series remote sensing and weather data, as well as static genotypic information. As a large number of features have been derived from the remote sensing data, feature importance analysis is conducted to identify and remove redundant features. A strategy to extract representative information from high-dimensional genetic markers is proposed. To enhance generalization and minimize the need for ground reference data, transfer learning strategies are proposed for selecting the most informative training samples from the target domain. Consequently, a pre-trained model can be refined with limited training samples. Field experiments were conducted over a sorghum breeding trial planted in multiple years with more than 600 testcross hybrids. The results show that the proposed LSTM-based RNN model can achieve high accuracies for single year prediction. Further, with the proposed transfer learning strategies, a pre-trained model can be refined with limited training samples from the target domain and predict biomass with an accuracy comparable to that from a trained-from-scratch model for both multiple experiments within a given year and across multiple years.

Predicting the Output of Solar Photovoltaic Panels in the Absence of Weather Data Using Only the Power Output of the Neighbouring Sites

There is an increasing need for capable models in the forecast of the output of solar photovoltaic panels. These models are vital for optimizing the performance and maintenance of PV systems. There is also a shortage of studies on forecasts of the output power of solar photovoltaics sites in the absence of meteorological data. Unlike common methods, this study explores numerous machine learning algorithms for forecasting the output of solar photovoltaic panels in the absence of weather data such as temperature, humidity and wind speed, which are often used when forecasting the output of solar PV panels. The considered models include Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Recurrent Neural Network (RNN) and Transformer. These models were used with the data collected from 50 different solar photo voltaic sites in South Korea, which consist of readings of the output of each of the sites collected at regular intervals. This study focuses on obtaining multistep forecasts for the multi-in multi-out, multi-in uni-out and uni-in uni-out settings. Detailed experimentation was carried out in each of these settings. Finally, for each of these settings and different lookback and forecast lengths, the best models were also identified.

Supervised Deep Learning Techniques for Image Description: A Systematic Review

Automatic image description, also known as image captioning, aims to describe the elements included in an image and their relationships. This task involves two research fields: computer vision and natural language processing; thus, it has received much attention in computer science. In this review paper, we follow the Kitchenham review methodology to present the most relevant approaches to image description methodologies based on deep learning. We focused on works using convolutional neural networks (CNN) to extract the characteristics of images and recurrent neural networks (RNN) for automatic sentence generation. As a result, 53 research articles using the encoder-decoder approach were selected, focusing only on supervised learning. The main contributions of this systematic review are: (i) to describe the most relevant image description papers implementing an encoder-decoder approach from 2014 to 2022 and (ii) to determine the main architectures, datasets, and metrics that have been applied to image description.

EMG-Based Estimation of Lower Limb Joint Angles and Moments Using Long Short-Term Memory Network

One of the fundamental limitations in human biomechanics is that we cannot directly obtain joint moments during natural movements without affecting the motion. However, estimating these values is feasible with inverse dynamics computation by employing external force plates, which can cover only a small area of the plate. This work investigated the Long Short-Term Memory (LSTM) network for the kinetics and kinematics prediction of human lower limbs when performing different activities without using force plates after the learning. We measured surface electromyography (sEMG) signals from 14 lower extremities muscles to generate a 112-dimensional input vector from three sets of features: root mean square, mean absolute value, and sixth-order autoregressive model coefficient parameters for each muscle in the LSTM network. With the recorded experimental data from the motion capture system and the force plates, human motions were reconstructed in a biomechanical simulation created using OpenSim v4.1, from which the joint kinematics and kinetics from left and right knees and ankles were retrieved to serve as output for training the LSTM. The estimation results using the LSTM model deviated from labels with average R² scores (knee angle: 97.25%, knee moment: 94.9%, ankle angle: 91.44%, and ankle moment: 85.44%). These results demonstrate the feasibility of the joint angle and moment estimation based solely on sEMG signals for multiple daily activities without requiring force plates and a motion capture system once the LSTM model is trained.

Neural learning rules for generating flexible predictions and computing the successor representation

The predictive nature of the hippocampus is thought to be useful for memory-guided cognitive behaviors. Inspired by the reinforcement learning literature, this notion has been formalized as a predictive map called the successor representation (SR). The SR captures a number of observations about hippocampal activity. However, the algorithm does not provide a neural mechanism for how such representations arise. Here, we show the dynamics of a recurrent neural network naturally calculate the SR when the synaptic weights match the transition probability matrix. Interestingly, the predictive horizon can be flexibly modulated simply by changing the network gain. We derive simple, biologically plausible learning rules to learn the SR in a recurrent network. We test our model with realistic inputs and match hippocampal data recorded during random foraging. Taken together, our results suggest that the SR is more accessible in neural circuits than previously thought and can support a broad range of cognitive functions.

Diagnosis of COVID-19 from X-rays using combined CNN-RNN architecture with transfer learning

Combating the COVID-19 pandemic has emerged as one of the most promising issues in global healthcare. Accurate and fast diagnosis of COVID-19 cases is required for the right medical treatment to control this pandemic. Chest radiography imaging techniques are more effective than the reverse-transcription polymerase chain reaction (RT-PCR) method in detecting coronavirus. Due to the limited availability of medical images, transfer learning is better suited to classify patterns in medical images. This paper presents a combined architecture of convolutional neural network (CNN) and recurrent neural network (RNN) to diagnose COVID-19 patients from chest X-rays. The deep transfer techniques used in this experiment are VGG19, DenseNet121, InceptionV3, and Inception-ResNetV2, where CNN is used to extract complex features from samples and classify them using RNN. In our experiments, the VGG19-RNN architecture outperformed all other networks in terms of accuracy. Finally, decision-making regions of images were visualized using gradient-weighted class activation mapping (Grad-CAM). The system achieved promising results compared to other existing systems and might be validated in the future when more samples would be available. The experiment demonstrated a good alternative method to diagnose COVID-19 for medical staff. 

All the data used during the study are openly available from the Mendeley data repository at https://data.mendeley.com/datasets/mxc6vb7svm. For further research, we have made the source code publicly available at https://github.com/Asraf047/COVID19-CNN-RNN.

Sensor Data Reconstruction for Dynamic Responses of Structures Using External Feedback of Recurrent Neural Network

An event of sensor faults in sensor networks deployed in structures might result in the degradation of the structural health monitoring system and lead to difficulties in structural condition assessment. Reconstruction techniques of the data for missing sensor channels were widely adopted to restore a dataset from all sensor channels. In this study, a recurrent neural network (RNN) model combined with external feedback is proposed to enhance the accuracy and effectiveness of sensor data reconstruction for measuring the dynamic responses of structures. The model utilizes spatial correlation rather than spatiotemporal correlation by explicitly feeding the previously reconstructed time series of defective sensor channels back to the input dataset. Because of the nature of spatial correlation, the proposed method generates robust and precise results regardless of the hyperparameters set in the RNN model. To verify the performance of the proposed method, simple RNN, long short-term memory, and gated recurrent unit models were trained using the acceleration datasets obtained from laboratory-scaled three- and six-story shear building frames.

Parameter estimation for WMTI-Watson model of white matter using encoder-decoder recurrent neural network

PURPOSE

Biophysical modeling of the diffusion MRI (dMRI) signal provides estimates of specific microstructural tissue properties. Although non-linear least squares (NLLS) is the most widespread fitting method, it suffers from local minima and high computational cost. Deep learning approaches are steadily replacing NLLS, but come with the limitation that the model needs to be retrained for each acquisition protocol and noise level. In this study, a novel fitting approach was proposed based on the encoder-decoder recurrent neural network (RNN) to accelerate model estimation with good generalization to various datasets.

METHODS

The white matter tract integrity (WMTI)-Watson model as an implementation of the Standard Model of diffusion in white matter derives its parameters indirectly from the diffusion and kurtosis tensors (DKI). The RNN-based solver, which estimates the WMTI-Watson model from DKI, is therefore more readily translatable to various data, irrespective of acquisition protocols as long as the DKI was pre-computed from the signal. An embedding approach was also used to render the model insensitive to potential differences in distributions between training data and experimental data. The analytical solution, NLLS, RNN-, and a multilayer perceptron (MLP)-based methods were evaluated on synthetic and in vivo datasets of rat and human brain.

RESULTS

The proposed RNN solver showed highly reduced computation time over the analytical solution and NLLS, with similar accuracy but improved robustness, and superior generalizability over MLP.

CONCLUSION

The RNN estimator can be easily applied to various datasets without retraining, which shows great potential for a widespread use.

Prediction of Blood Risk Score in Diabetes Using Deep Neural Networks

Improving the prediction of blood glucose concentration may improve the quality of life of people living with type 1 diabetes by enabling them to better manage their care. Given the anticipated benefits of such a prediction, numerous methods have been proposed. Rather than attempting to predict glucose concentration, a deep learning framework for prediction is proposed in which prediction is performed using a scale for hypo- and hyper-glycemia risk. Using the blood glucose risk score formula proposed by Kovatchev et al., models with different architectures were trained, including, a recurrent neural network (RNN), a gated recurrent unit (GRU), a long short-term memory (LSTM) network, and an encoder-like convolutional neural network (CNN). The models were trained using the OpenAPS Data Commons data set, comprising 139 individuals, each with tens of thousands of continuous glucose monitor (CGM) data points. The training set was composed of 7% of the data set, while the remaining was used for testing. Performance comparisons between the different architectures are presented and discussed. To evaluate these predictions, performance results are compared with the last measurement (LM) prediction, through a sample-and-hold approach continuing the last known measurement forward. The results obtained are competitive when compared to other deep learning methods. A root mean squared error (RMSE) of 16 mg/dL, 24 mg/dL, and 37 mg/dL were obtained for CNN prediction horizons of 15, 30, and 60 min, respectively. However, no significant improvements were found for the deep learning models compared to LM prediction. Performance was found to be highly dependent on architecture and the prediction horizon. Lastly, a metric to assess model performance by weighing each prediction point error with the corresponding blood glucose risk score is proposed. Two main conclusions are drawn. Firstly, going forward, there is a need to benchmark model performance using LM prediction to enable the comparison between results obtained from different data sets. Secondly, model-agnostic data-driven deep learning models may only be meaningful when combined with mechanistic physiological models; here, it is argued that neural ordinary differential equations may combine the best of both approaches. These findings are based on the OpenAPS Data Commons data set and are to be validated in other independent data sets.

Artificial intelligence techniques for retinal prostheses: a comprehensive review and future direction

Objective. Retinal prostheses are promising devices to restore vision for patients with severe age-related macular degeneration or retinitis pigmentosa disease. The visual processing mechanism embodied in retinal prostheses play an important role in the restoration effect. Its performance depends on our understanding of the retina's working mechanism and the evolvement of computer vision models. Recently, remarkable progress has been made in the field of processing algorithm for retinal prostheses where the new discovery of the retina's working principle and state-of-the-arts computer vision models are combined together.Approach. We investigated the related research on artificial intelligence techniques for retinal prostheses. The processing algorithm in these studies could be attributed to three types: computer vision-related methods, biophysical models, and deep learning models.Main results. In this review, we first illustrate the structure and function of the normal and degenerated retina, then demonstrate the vision rehabilitation mechanism of three representative retinal prostheses. It is necessary to summarize the computational frameworks abstracted from the normal retina. In addition, the development and feature of three types of different processing algorithms are summarized. Finally, we analyze the bottleneck in existing algorithms and propose our prospect about the future directions to improve the restoration effect.Significance. This review systematically summarizes existing processing models for predicting the response of the retina to external stimuli. What's more, the suggestions for future direction may inspire researchers in this field to design better algorithms for retinal prostheses.

CNN-RNN Network Integration for the Diagnosis of COVID-19 Using Chest X-ray and CT Images

The 2019 coronavirus disease (COVID-19) has rapidly spread across the globe. It is crucial to identify positive cases as rapidly as humanely possible to provide appropriate treatment for patients and prevent the pandemic from spreading further. Both chest X-ray and computed tomography (CT) images are capable of accurately diagnosing COVID-19. To distinguish lung illnesses (i.e., COVID-19 and pneumonia) from normal cases using chest X-ray and CT images, we combined convolutional neural network (CNN) and recurrent neural network (RNN) models by replacing the fully connected layers of CNN with a version of RNN. In this framework, the attributes of CNNs were utilized to extract features and those of RNNs to calculate dependencies and classification base on extracted features. CNN models VGG19, ResNet152V2, and DenseNet121 were combined with long short-term memory (LSTM) and gated recurrent unit (GRU) RNN models, which are convenient to develop because these networks are all available as features on many platforms. The proposed method is evaluated using a large dataset totaling 16,210 X-ray and CT images (5252 COVID-19 images, 6154 pneumonia images, and 4804 normal images) were taken from several databases, which had various image sizes, brightness levels, and viewing angles. Their image quality was enhanced via normalization, gamma correction, and contrast-limited adaptive histogram equalization. The ResNet152V2 with GRU model achieved the best architecture with an accuracy of 93.37%, an F1 score of 93.54%, a precision of 93.73%, and a recall of 93.47%. From the experimental results, the proposed method is highly effective in distinguishing lung diseases. Furthermore, both CT and X-ray images can be used as input for classification, allowing for the rapid and easy detection of COVID-19.