Artificial Neural Model Research Articles

Artificial Neural Network Prediction of Mortality in Cancer Patients Presenting for Radiation Therapy at a Multisite Institution.

For many decades, the management of cancer has utilized radiation therapy, which continues to evolve with technology to improve patient outcomes. However, despite the standardization of treatment plans and the establishment of best clinical practices based on prospective, randomized trials and adherence to National Comprehensive Cancer Network (NCCN) guidelines, the outcomes from radiation therapy are highly variable and dependent on a number of factors, including patient demographics, tumor characteristics/histology, and treatment parameters. In this study, we attempt to use available patient data and treatment parameters at the time of radiation therapy to predict future outcomes using artificial intelligence (AI). Six thousand five hundred ninety-five cases of patients who completed radiation treatment were selected retrospectively and used to train artificial neural networks (ANNs) and baseline models (i.e., logistic regression, random forest, support vector machines [SVMs], gradient boosting [XGBoost]) for binary classification of mortality at multiple time points ranging from six months to five years post-treatment. A hyperparameter grid search was used to identify the optimal network architecture for each time point, using sensitivity as the primary outcome metric. The median age was 75 years (range: 2-102 years). There were 63.8% females and 36.1% males. The results indicate that ANNs were able to successfully perform binary mortality prediction with an accuracy greater than random chance and greater sensitivity than baseline models used. The best-performing algorithm was the ANN, which achieved a sensitivity of 83.00% ± 4.89% for five-year mortality. The neural network was able to achieve higher sensitivity than Logistic Regression, SVM Random Forest, and XGBoost across all output target variables, demonstrating the utility of a neural network model for mortality prediction on the provided dataset.

Robust optimization of a novel ultraviolet (UV) photoreactor for water disinfection: A neural network approach

To optimize the ultraviolet (UV) water disinfection process, it is crucial to determine the ideal geometric dimensions of a corresponding model that enhance performance while minimizing the impact of uncertain photoreactor inputs. As water treatment directly affects people's lives, it is crucial to eliminate the risks associated with the non-ideal performance of disinfection photoreactors. Input uncertainties greatly affect photoreactor performance, making it essential to develop a robust optimization algorithm in advance to mitigate these effects and minimize the physical and financial resources required for constructing the photoreactors. In the suggested algorithm, a two-objective genetic algorithm is integrated with a non-intrusive polynomial chaos expansion (PCE) technique. Additionally, the Sobol sampling method is employed to select the necessary samples for understanding the system's behavior. An artificial neural network surrogate model is trained using sufficient data points derived from computational fluid dynamics (CFD) simulations. A novel type of UV photoreactors working based on exterior reflectors is chosen to optimize the process with three uncertain input parameters, including UV lamp power, UV transmittance of water, and diffusive fraction of the reflective surface. In addition, four geometrical design variables are considered to find the optimal configuration of the photoreactor. The standard deviation (SD) and the reciprocal of log reduction value (LRV) are set as the objective functions, calculated using PCE. The optimal design provides a LRV of 3.95 with SD of 0.2. The coefficient of variation (CoV) of the model significantly declines up to 7%, indicating the decreased sensitivity of the photoreactor to the input uncertainties. Additionally, it is discovered that the robust model exhibits minimal sensitivity to changes in reflectivity in various flow rates, and its output variability aligns with the SD obtained through robust optimization.

Comparison Study on Modelling and Prediction of Weather Parameters Combining Exponential Smoothing and Artificial Neural Network Models in Different Zones of Kerala

Prediction of weather parameters with maximum precision is the main objective faced by different climatologists all over the world. The researchers devised a number of ways for performing weather parameter predictions. For modelling and prediction of weather parameters, exponential smoothing (ETS), artificial neural networks (ANN) and ETS-ANN hybrid models were employed for the northern, central and southern zones of Kerala. The weather parameters used in the study are relative humidity and wind speed. The monthly data were collected for the northern and central zones, for a period of 39 years (1982-2020), whereas for the southern zone it was for a period of 36 years (1985-2020). The result suggested that the best fitted model for relative humidity was the ANN for the central and southern zones whereas for northern zone it was the ETS model. The accuracy of the model was calculated using MSE, RMSE, MAE, MAPE and R2 values. The ETS performed best in all three zones of Kerala when it came to wind speed. The study also advocated that traditional model ETS and ANN is performing better compared to the newly developed hybrid model in prediction of weather parameters. The best model chosen for weather parameters of three different zones are used to forecast for the next five years.

Hydrodynamic shape optimization of an autonomous and remotely-operated vehicle via a multi-surrogate model

The design requirements for the hydrodynamic performance of underwater vehicles vary significantly depending on the application. Optimization without an initial model results in challenges such as large design domains, nonlinear complexity, and high data requirements. To optimize the hydrodynamic performance during the design process, in this paper, a multi-surrogate model was employed to progressively design the shape of an autonomous and remotely-operated vehicle. Based on the data characteristics of different stages, this approach strikes a balance among data quantity, prediction accuracy, and multi-objective requirements. An artificial neural network prediction surrogate model was constructed based on the principle of the minimum prediction factor using the optimal Latin hypercube sampling method. During the optimization of the design domain, the optimum objective was to minimize the dimensionless force (G(x)). During the multi-objective optimization stage, a regression kriging surrogate model was constructed based on a support vector product. The optimization objectives were to maximize volume and minimize dimensionless forces. This enabled the overall design process to attain optimal Pareto solutions within the design domain while simultaneously ensuring high prediction accuracy and minimum data requirements. The results obtained are consistent with the simulation comparison, thus verifying the reliability of the entire optimization process.

Exploratory analysis of new craniometric measures for the investigation of biological sex using open-access statistical and machine-learning tools on a cone-beam computed tomography sample.

Investigation of the biological sex of human remains is a crucial aspect of physical anthropology. However, due to varying states of skeletal preservation, multiple approaches and structures of interest need to be explored. This research aims to investigate the potential use of distances between bifrontal breadth (FMB), infraorbital foramina distance (IOD), nasal breadth (NLB), inter-canine width (ICD), and distance between mental foramina (MFD) for combined sex prediction through traditional statistical methods and through open-access machine-learning tools. Ethical approval was obtained from the ethics committee, and out of 100 cone beam computed tomography (CBCT) scans, 54 individuals were selected with all the points visible. Ten extra exams were chosen to test the predictors developed from the learning sample. Descriptive analysis of measurements, standard deviation, and standard error were obtained. T-student and Mann-Whitney tests were utilized to assess the sex differences within the variables. A logistic regression equation was developed and tested for the investigation of the biological sex as well as decision trees, random forest, and artificial neural networks machine-learning models. The results indicate a strong correlation between the measurements and the sex of individuals. When combined, the measurements were able to predict sex using a regression formula or machine learning based models which can be exported and added to software or webpages. Considering the methods, the estimations showed an accuracy rate superior to 80% for males and 82% for females. All skulls in the test sample were accurately predicted by both statistical and machine-learning models. This exploratory study successfully established a correlation between facial measurements and the sex of individuals, validating the prediction potential of machine learning, augmenting the investigative tools available to experts with a high differentiation potential.

Wavelet–ANN hybrid model evaluation in seepage prediction in nonhomogeneous earthen dams

ABSTRACT In this study, novel methods such as wavelet–artificial neural network hybrid models and artificial neural network models were used to predict seepage from the Zonouz earthen dam. The dataset consisted of 972 piezometric data points. Statistical fitting methods such as root mean squared error, determination coefficient, scatter plots, and data distribution diagrams were used to evaluate the results. The findings indicated that the wavelet–artificial neural network hybrid model was more accurate than the artificial neural network model. Specifically, during training, the wavelet–artificial neural network hybrid model had determination coefficients and root mean squared errors of 0.820, 0.680, 743.39, and 792.52, while the artificial neural network model had 0.700, 0.600, 426.39, and 131.45. Similarly, during validation, the wavelet–artificial neural network hybrid model had determination coefficients and root mean squared errors of 0.700, 0.600, 426.39, and 131.45, while the artificial neural network model had 0.823, 0.680, 743.39, and 792.52. Therefore, the wavelet–artificial neural network hybrid model can be proposed as a precise method for predicting seepage in earthen dams and is more accurate than the artificial neural network model. This study highlights the importance of preventing dam failures and using advanced modeling techniques for better predictions and preventive measures.

Deep artificial neural network-powered phase field model for predicting damage characteristic in brittle composite under varying configurations

This work introduces a novel artificial neural network (ANN)-powered phase field model, offering rapid and precise predictions of fracture propagation in brittle materials. To improve the capabilities of the ANN model, we incorporate a loop of conditions into its core to regulate the absolute percentage error for each observation point, that filters and consistently selects the most accurate outcome. This algorithm enables our model to better adapt to the highly sensitive validation data arising from varying configurations. The effectiveness of the approach is illustrated through three examples involving changes in the microgeometry and material properties of steel fiber-reinforced high-strength concrete structures. Indeed, the predicted outcomes from the improved ANN phase field model in terms of stress–strain relationship, and crack propagation path demonstrates an outperformance compared with that based on the extreme gradient boosting method, a leading regression machine learning technique for tabular data. Additionally, the introduced model exhibits a remarkable speed advantage, being 180 times faster than traditional phase field simulations, and provides results at nearly any fiber location, demonstrating superiority over the phase field model. This study marks a significant advancement in the application of artificial intelligence for accurately predicting crack propagation paths in composite materials, particularly in cases involving the relative positioning of the fiber and initial crack location.

An artificial neural network diagnostic model for scleroderma and immune cell infiltration analysis based on mitochondria-associated genes

To establish a diagnostic model for scleroderma by combining machine learning and artificial neural network based on mitochondria-related genes. The GSE95065 and GSE59785 datasets of scleroderma from GEO database were used for analyzing expressions of mitochondria-related genes, and the differential genes were identified by Random forest, LASSO regression and SVM algorithms. Based on these differential genes, an artificial neural network model was constructed, and its diagnostic accuracy was evaluated by 10-fold crossover verification and ROC curve analysis using the verification dataset GSE76807. The mRNA expressions of the key genes were verified by RT-qPCR in a mouse model of scleroderma. The CIBERSORT algorithm was used to estimate the bioinformatic association between scleroderma and the screened biomarkers. A total of 24 differential genes were obtained, including 11 up-regulated and 13 down-regulated genes. Seven most relevant mitochondria-related genes (POLB, GSR, KRAS, NT5DC2, NOX4, IGF1, and TGM2) were screened using 3 machine learning algorithms, and the artificial neural network diagnostic model was constructed. The model showed an area under the ROC curves of 0.984 for scleroderma diagnosis (0.740 for the verification dataset and 0.980 for cross-over validation). RT-qPCR detected significant up-regulation of POLB, GSR, KRAS, NOX4, IGF1 and TGM2 mRNAs and significant down-regulation of NT5DC2 in the mouse models of scleroderma. Immune cell infiltration analysis showed that the differential genes in scleroderma were associated with follicular helper T cells, immature B cells, resting dendritic cells, memory activated CD4+T cells, M0 macrophages, monocytes, resting memory CD4+T cells and mast cell activation. The artificial neural network diagnostic model for scleroderma established in this study provides a new perspective for exploring the pathogenesis of scleroderma.

Assessing bioenergy prospects of algal biomass and yard waste using an integrated hydrothermal carbonization and pyrolysis (HTC–PY): A detailed emission–to–ash characterization via diverse hyphenated analytical techniques and modelling strategies

Hydrothermal carbonization (HTC) presents a promising method for converting carbonaceous waste into renewable fuels, facilitating energy recovery. This study focuses on the HTC pretreatment of microalgae (Chlorella sorokiniana) and yard waste, examining their physicochemical structural changes, component migration, and potential for valorization via integrated HTC and pyrolysis processes. HTC notably enhances the carbon content while reducing oxygen and nitrogen levels in these feedstocks, producing a clean solid fuel with high energy density. The study investigates the pyrolytic behavior of hydrochars and their blends using different analytical techniques, kinetic modeling, and fixed–bed reactor experiments. The blends with 90% and 70% of microalgae hydrochar (CSH) exhibited the most significant synergistic effects during co–pyrolysis, evident in its devolatilization behaviors and reduced activation energies. Integrating artificial neural networks and response surface methodology models facilitated optimization of pyrolysis parameters and accurate prediction of bio–oil, biochar, and gas yields. The analytical techniques enabled characterization of pyrolysis products, revealing notable changes in bio–oil and biochar compositions, suggesting potential enhancements. The study comprehensively evaluates the HTC–pyrolysis approach’s potential for converting microalgae and lignocellulosic biomass into eco–friendly biofuels for industrial applications.

Development of an artificial neural network based-prediction model for bond strength of FRP bars in concrete

Fiber-reinforced polymer (FRP) bars have garnered increasing attention in recent years due to their superior corrosion resistance, offering a potential solution to the significant drawback of steel corrosion in concrete. For the widespread utilization of FRP bars in concrete structures, determining the bond strength between FRP bars and concrete is a crucial topic. This study seeks to develop a prediction model to estimate the bond strength of FRP bars in concrete, utilizing an extended dataset from 1010 pull-out tests. Initially, the study evaluates the applicability of several bond strength formulas from existing codes. Subsequently, two prediction models, namely a multivariate linear regression model and an artificial neural network (ANN) model, are introduced for estimating the bond strength of FRP bars in concrete. The results indicate that the correlation between the evaluation values of existing formulas and the experimental value is very low. This is because these formulas have not yet been updated to encompass the expanded usage scopes of FRP bars with various surface processing methods and types of concrete. While the multivariate linear regression model outperforms these formulas, its accuracy is still relatively low; in contrast, the ANN demonstrates superior performance, achieving an R^2 value for both the validation and test set of more than 0.92. The findings highlight that, when considering a broader range of applications, the ANN serves as a robust tool for accurately predicting the bond strength of FRP bars in concrete, in comparison to traditional formulas and linear regression models. This assessment approach provides engineers with a convenient, high-precision tool for designs utilizing various forms of FRP bars and diverse types of concrete in practical design scenarios

Compound-choking theory and artificial neural networks-based hybrid modeling for supersonic ejectors

As the core component of ejector refrigeration systems, the research of ejector mechanisms has always been a hot topic. Based on the compound-choking theory, this paper studies the relationship between the entrainment ratio and critical back pressure and designs a neural network-based online error compensation mechanism for supersonic ejectors. Firstly, a thermodynamic model of the ejector is developed by combining the ideal gas model, the compound-choking theory, and the constant-pressure mixing theory. Then, the thermodynamic relationship between the entrainment ratio and critical back pressure is derived by introducing the idea of a hypothetical mixing cross section. Compared with the existing results, our method is able to improve critical back pressure prediction accuracy by virtue of the actual entrainment ratio. Therefore, a new adaptive error compensation algorithm is proposed for supersonic ejectors based on neural networks. Thus, the proposed scheme can be regarded as a hybrid modeling method for supersonic ejectors, which combines thermodynamic laws and data-driven artificial intelligence algorithms. Experimental data from ejector refrigeration systems with R141b, R245fa, water vapor and R134a are used in this paper to verify the validity of the model. The simulation results show that our method can effectively predict the critical back pressure, and all the prediction errors are less than 10%.

Clinical Impact of an AI Decision Support System for Detection of Intracranial Hemorrhage in CT Scans.

This study aimed to evaluate the predictive value and clinical impact of a clinically implemented artificial neural network software model. The software detects intracranial hemorrhage (ICH) from head computed tomography (CT) scans and artificial intelligence (AI)-identified positive cases are then annotated in the work list for early radiologist evaluation. The index test was AI detection by the program Zebra Medical Vision-HealthICH+. Radiologist-confirmed ICH was the reference standard. The study compared whether time benefits from using the AI model led to faster escalation of patient care or surgery within the first 24 h. A total of 2,306 patients were evaluated by the software, and 288 AI-positive cases were included. The AI tool had a positive predictive value of 0.823. There was, however, no significant time reduction when comparing the patients who required escalation of care and those who did not. There was also no significant time reduction in those who required acute surgery compared with those who did not. Among the individual patients with reduced time delay, no cases with evident clinical benefit were identified. Although the clinically implemented AI-based decision support system showed adequate predictive value in identifying ICH, there was no significant clinical benefit for the patients in our setting. While AI-assisted detection of ICH shows great promise from a technical perspective, there remains a need to evaluate the clinical impact and perform external validation across different settings.

Improved Artificial Neural Network Classification Model based Metaheuristic Optimization for Handwritten Character Recognition

This study addresses the concerns regarding the performance of Handwritten Character Recognition (HCR) systems, focusing on the classification stage. It is widely acknowledged that the development of the classification model significantly impacts the overall performance of HCR. The problems identified specifically pertain to the classification model, particularly in the context of the Artificial Neural Network (ANN) learning problem, leading to low accuracy in recognizing handwritten characters. The objective of this study is to improve and refine the ANN classification model to achieve better HCR. To achieve this goal, this study proposed a hybrid Flower Pollination Algorithm with Artificial Neural Network (FPA-ANN) classification model for HCR. The FPA is one of the metaheuristic approaches is utilized as an optimization technique to enhance the performance of ANN, particularly by optimizing the network training process of ANN. The experimentation phase involves using the National Institute of Standards and Technology (NIST) handwritten character database. Finally, the proposed FPA-ANN classification model is analysed based on generated confusion matrix and evaluated performance of the classification model in terms of precision, sensitivity, specificity, F-score and accuracy.

Determination of hardness and Young's modulus in fcc Cu–Ni–Sn–Al alloys via high-throughput experiments, CALPHAD approach and machine learning

Hardness and Young's modulus are critical indicators in the design of innovative Cu–Ni–Sn–Al alloys with desired elastic and strength properties. In this study, the composition-dependent hardness and Young's modulus in the fcc Cu–Ni–Sn–Al alloys were determined using high-throughput experiments, the CALPHAD (CALculation of PHAse Diagrams) approach, and machine learning (ML) model. By combining the diffusion-couples tool, nanoindentation, and electron probe microanalysis (EPMA) techniques, the 341 sets of composition-related hardness and Young's modulus in the fcc Cu–Ni–Sn–Al system, as well as its sub-binary and sub-ternary systems, were effectively determined. These measured data were subsequently utilized to establish the relationship of Young's modulus and hardness with respect to composition in the fcc Cu–Ni–Sn–Al system through the combined application of CALPHAD and back propagation (BP) artificial neural network machine learning models. A comparison between the two approaches revealed that the ML method achieves higher accuracy in ternary and quaternary alloys compared to the CALPHAD method. The impact of alloying elements on the values of hardness and Young's modulus was analyzed, demonstrating that the Ni content has the largest effect on hardness, while the Young's modulus increases with addition of Ni but decreases with the addition of Al and Sn. Finally, the effects of solute elements on the nanomechanical properties of aged Cu1-x(Ni3Sn)x and Cu–9Ni–6Sn-xAl alloys were examined.

Habitat Quality Analysis and Future Simulation Based on Artificial Neural Network-cellular Automata Model

Habitat Quality Analysis and Future Simulation Based on Artificial Neural Network-cellular Automata Model

A Simple Artificial Neural Model to Predict Dambovita River Temperature Affected by Urban Heat Islands in Bucharest City

Water bodies can offer local microclimates that have the potential to attenuate the effects of urban heat islands by reducing local temperature. This capability is shaded when the river is channelized. In such cases, the river temperature rises during hot periods, leading to negative impacts on the water quality. The main aim of this paper is to develop a local simple model to predict the temperature of the Dâmbovița River at its exit from Bucharest City, the capital of Romania. The location is chosen based on the historical critical impacts, in terms of extreme heatwaves that took place during hot summers, as well as future possible risks due to climate change. The water temperature prediction model is based on an artificial neural network that uses the Levenberg–Marquardt algorithm, due to its stability and rapid convergence capabilities. The model forecasts, with an accuracy of ±1 °C, the water temperature in an ungauged, downstream location, as a function of measured air and upstream water temperatures. The proposed model represents a first attempt to provide water managers in Bucharest City with a useful tool that will allow them to take timely measures to counteract the unwanted effects that can be generated by high water temperatures.

An artificial neural network-source apportionment-based prediction model for carbon monoxide from total number of ships calling by ports in Malaysia

An artificial neural network-source apportionment-based prediction model for carbon monoxide from total number of ships calling by ports in Malaysia

Measuring Color Index of Transformer Oil–Enabling Single-Wavelength Spectroscopy With Artificial Neural Network-Fuzzy Logic Model

Conventionally, the color index of transformer oil is determined by a color comparator based on the American Society for Testing and Materials (ASTM) D 1500 standard. The equipment requires humans to operate, which leads to human error and limited number of samples tested per day. This work proposes the utilization of single-wavelength spectroscopy with 405 nm laser diode using artificial neural network (ANN) to determine the color index of transformer oil. Two ANN models were developed using data collected from 50 oil samples with different optical pathlengths of 1 to 10 mm, and laser output powers of 1 to 15 mW. The first model classified the input into different color indices and another model correlated the input parameters through regression analysis to determine the color index. A hybrid ANN-fuzzy logic model was also developed to improve the color index prediction. The root-mean-squared error (RMSE) obtained from the prediction by ANN regressor and ANN classifier are 0.5602 and 0.6416, respectively. The hybrid ANN-fuzzy logic model improves the RMSE especially for optical pathlengths <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$&lt;$</tex-math> </inline-formula> 5 mm, which is required for measuring samples with high color index. This proposed method reduces the dependency on complex optoelectronic hardware to obtain highly accurate results. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —Unlike the conventional testing method for color index of transformer oil that requires human observation, the findings of this study enables the possibility of compact and smart portable device through the utilization of single wavelength spectroscopy with machine learning models. With no human involvement, more computational power with lesser hardware dependency, the maintenance cost and error can be reduced. This proposed method can potentially be applied to measure the color of other amber-colored liquid products such as olive oil, honey and others.

Artificial neural network-substituted transition model for crossflow instability: Modeling strategy and application prospect

Data-driven approaches have made preliminary inroads into the area of transition–turbulence modeling, but are still in their infancy with regard to widespread industrial adoption. This paper establishes an artificial neural network (ANN)-based transition model to enhance the capacity of capturing the crossflow (CF) transition phenomena, which are frequently identified over a wide range of aerodynamic problems. By taking a new CF-extended shear stress transport (SST) transition-predictive (SST-γ) model as the baseline, a mapping from mean flow variables to transition intermittency factor (γ) is constructed by ANN algorithm at various Mach and Reynolds numbers of an infinite swept wing. Generalizability of the resulting ANN-based (SST-γANN) model is fully validated in the same infinite swept wing, an inclined 6:1 prolate spheroid, and a finite swept wing in extensive experiment regimes, together with two effective a priori analysis strategies. Furthermore, the calculation efficiency, grid dependence, and performance of the present model in non-typical transitional flow are also assessed to inspect its industrial feasibility, followed by the elucidation of rationality behind the preliminary success and transferability of present framework. The results manifest that the SST-γANN model aligns well with the benchmark SST-γ model, and both can capture the CF transition accurately compared with their experiment counterpart, completely breaking through the disability of original SST-γ model without CF correction. In addition, good properties of efficiency, robustness, and generalizability are achieved for the ANN-alternative transition model, together with the usability of present framework across various transitional flows.

Artificial Neural Network Language Models Predict Human Brain Responses to Language Even After a Developmentally Realistic Amount of Training.

Artificial neural networks have emerged as computationally plausible models of human language processing. A major criticism of these models is that the amount of training data they receive far exceeds that of humans during language learning. Here, we use two complementary approaches to ask how the models' ability to capture human fMRI responses to sentences is affected by the amount of training data. First, we evaluate GPT-2 models trained on 1 million, 10 million, 100 million, or 1 billion words against an fMRI benchmark. We consider the 100-million-word model to be developmentally plausible in terms of the amount of training data given that this amount is similar to what children are estimated to be exposed to during the first 10 years of life. Second, we test the performance of a GPT-2 model trained on a 9-billion-token dataset to reach state-of-the-art next-word prediction performance on the human benchmark at different stages during training. Across both approaches, we find that (i) the models trained on a developmentally plausible amount of data already achieve near-maximal performance in capturing fMRI responses to sentences. Further, (ii) lower perplexity-a measure of next-word prediction performance-is associated with stronger alignment with human data, suggesting that models that have received enough training to achieve sufficiently high next-word prediction performance also acquire representations of sentences that are predictive of human fMRI responses. In tandem, these findings establish that although some training is necessary for the models' predictive ability, a developmentally realistic amount of training (∼100 million words) may suffice.