Robust optimization of convolutional neural networks with a uniform experiment design method: a case of phonocardiogram testing in patients with heart diseases.

BACKGROUND: Heart sound measurement is crucial for analyzing and diagnosing patients with heart diseases. This study employed phonocardiogram signals as the input signal for heart disease analysis due to the accessibility of the respective method. This study referenced preprocessing techniques proposed by other researchers for the conversion of phonocardiogram signals into characteristic images composed using frequency subband. Image recognition was then conducted through the use of convolutional neural networks (CNNs), in order to classify the predicted of phonocardiogram signals as normal or abnormal. However, CNN requires the tuning of multiple hyperparameters, which entails an optimization problem for the hyperparameters in the model. To maximize CNN robustness, the uniform experiment design method and a science-based methodical experiment design were used to optimize CNN hyperparameters in this study. RESULTS: An artificial intelligence prediction model was constructed using CNN, and the uniform experiment design method was proposed to acquire hyperparameters for optimal CNN robustness. The results indicate Filters ([Formula: see text]), Stride ([Formula: see text]), Activation functions ([Formula: see text]), and Dropout ([Formula: see text]) to be significant factors considerably influencing the ability of CNN to distinguish among heart sound states. Finally, the confirmation experiment was conducted, and the hyperparameter combination for optimal model robustness was Filters ([Formula: see text]) = 32, Kernel Size ([Formula: see text] = 3 × 3, Stride ([Formula: see text]) = (1,1), Padding ([Formula: see text] as same, Optimizer ([Formula: see text] as the stochastic gradient descent, Activation functions ([Formula: see text]) as relu, and Dropout ([Formula: see text]) = 0.544. With this combination of parameters, the model had an average prediction accuracy rate of 0.787 and standard deviation of 0. CONCLUSION: In this study, phonocardiogram signals were used for the early prediction of heart diseases. The science-based and methodical uniform experiment design was used for the optimization of CNN hyperparameters to construct a CNN with optimal robustness. The results revealed that the constructed model exhibited robustness and an acceptable accuracy rate. Other literature has failed to address hyperparameter optimization problems in CNN; a method is subsequently proposed for robust CNN optimization, thereby solving this problem.

Artificial intelligence classification model for macular degeneration images: a robust optimization framework for residual neural networks.

BACKGROUND: The prevalence of chronic disease is growing in aging societies, and artificial-intelligence-assisted interpretation of macular degeneration images is a topic that merits research. This study proposes a residual neural network (ResNet) model constructed using uniform design. The ResNet model is an artificial intelligence model that classifies macular degeneration images and can assist medical professionals in related tests and classification tasks, enhance confidence in making diagnoses, and reassure patients. However, the various hyperparameters in a ResNet lead to the problem of hyperparameter optimization in the model. This study employed uniform design-a systematic, scientific experimental design-to optimize the hyperparameters of the ResNet and establish a ResNet with optimal robustness. RESULTS: An open dataset of macular degeneration images ( https://data.mendeley.com/datasets/rscbjbr9sj/3 ) was divided into training, validation, and test datasets. According to accuracy, false negative rate, and signal-to-noise ratio, this study used uniform design to determine the optimal combination of ResNet hyperparameters. The ResNet model was tested and the results compared with results obtained in a previous study using the same dataset. The ResNet model achieved higher optimal accuracy (0.9907), higher mean accuracy (0.9848), and a lower mean false negative rate (0.015) than did the model previously reported. The optimal ResNet hyperparameter combination identified using the uniform design method exhibited excellent performance. CONCLUSION: The high stability of the ResNet model established using uniform design is attributable to the study's strict focus on achieving both high accuracy and low standard deviation. This study optimized the hyperparameters of the ResNet model by using uniform design because the design features uniform distribution of experimental points and facilitates effective determination of the representative parameter combination, reducing the time required for parameter design and fulfilling the requirements of a systematic parameter design process.

Integrating Taguchi Method and Gray Relational Analysis for Auto Locks by Using Multiobjective Design in Computer-Aided Engineering.

In automobiles, lock parts are matched with inserts, and this is a crucial quality standard for the dimensional accuracy of the molding. This study employed moldflow analysis to explore the influence of various injection molding process parameters on the warpage deformation. Deformation of the plastic part is caused by the nonuniform product temperature distribution in the manufacturing process. Furthermore, improper parameter design leads to substantial warpage and deformation. The Taguchi robust design method and gray correlation analysis were used to optimize the process parameters. Multiobjective quality analysis was performed for achieving a uniform temperature distribution and reducing the warpage deformation to obtain the optimal injection molding process parameters. Subsequently, three water cooling system designs-original cooling, U-shaped cooling, and conformal cooling-were tested to modify the temperature distribution and reduce the warpage. Taguchi gray correlation analysis revealed that the main influencing parameter was the mold temperature followed by the holding pressure. Moreover, the results indicated that the conformal cooling system improved the average temperature distribution.

Application of Intelligent Modeling Method to Optimize the Multiple Quality Characteristics of the Injection Molding Process of Automobile Lock Parts.

This study focuses on applying intelligent modeling methods to different injection molding process parameters, to analyze the influence of temperature distribution and warpage on the actual development of auto locks. It explores the auto locks using computer-aided engineering (CAE) simulation performance analysis and the optimization of process parameters by combining multiple quality characteristics (warpage and average temperature). In this experimental design, combinations were explored for each single objective optimization process parameter, using the Taguchi robust design process, with the L18 (21 × 37) orthogonal table. The control factors were injection time, material temperature, mold temperature, injection pressure, packing pressure, packing time, cooling liquid, and cooling temperature. The warpage and temperature distribution were analysed as performance indices. Then, signal-to-noise ratios (S/N ratios) were calculated. Gray correlation analysis, with normalization of the S/N ratio, was used to obtain the gray correlation coefficient, which was substituted into the fuzzy theory to obtain the multiple performance characteristic index. The maximum multiple performance characteristic index was used to find multiple quality characteristic-optimized process parameters. The optimal injection molding process parameters with single objective are a warpage of 0.783 mm and an average temperature of 235.23 °C. The optimal parameters with multi-objective are a warpage of 0.753 mm and an average temperature of 238.71 °C. The optimal parameters were then used to explore the different cooling designs (original cooling, square cooling, and conformal cooling), considering the effect of the plastics temperature distribution and warpage. The results showed that, based on the design of the different cooling systems, conformal cooling obtained an optimal warpage of 0.661 mm and a temperature of 237.62 °C. Furthermore, the conformal cooling system is smaller than the original cooling system; it reduces the warpage by 12.2%, and the average temperature by 0.46%.

Adaptive neuro-fuzzy approach for predicting hardness of deposited TiN/ZrN multilayer coatings.

This paper presents an adaptive neuro-fuzzy approach based on first order function of fuzzy model for establishing the relationship between control factors and thin films properties of TiN/ZrN coatings on Si(100) wafer substrates. A statistical model was designed to explore the space of the processes by an orthogonal array scheme. Eight control factors of closed unbalance magnetron sputtering system were selected for modeling the process, such as interlayer material, argon and nitrogen flow rate, titanium and zirconium target current, rotation speed, work distance, and bias voltage. Analysis of variance (ANOVA) was carried out for determining the influence of control factors. In this study, with the application of ANOVA, the smallest effect of control factors was eliminated. The adaptive neuro-fuzzy inference system (ANFIS) was applied as a tool to model the deposited process with five significant control factors. The experimental results show that ANFIS demonstrates better accuracy than additive model for the film hardness. The root mean square error between prediction values and experimental values were archived to 0.04.

Tuning the structure and parameters of a neural network by using hybrid Taguchi-genetic algorithm.

In this paper, a hybrid Taguchi-genetic algorithm (HTGA) is applied to solve the problem of tuning both network structure and parameters of a feedforward neural network. The HTGA approach is a method of combining the traditional genetic algorithm (TGA), which has a powerful global exploration capability, with the Taguchi method, which can exploit the optimum offspring. The Taguchi method is inserted between crossover and mutation operations of a TGA. Then, the systematic reasoning ability of the Taguchi method is incorporated in the crossover operations to select the better genes to achieve crossover, and consequently enhance the genetic algorithms. Therefore, the HTGA approach can be more robust, statistically sound, and quickly convergent. First, the authors evaluate the performance of the presented HTGA approach by studying some global numerical optimization problems. Then, the presented HTGA approach is effectively applied to solve three examples on forecasting the sunspot numbers, tuning the associative memory, and solving the XOR problem. The numbers of hidden nodes and the links of the feedforward neural network are chosen by increasing them from small numbers until the learning performance is good enough. As a result, a partially connected feedforward neural network can be obtained after tuning. This implies that the cost of implementation of the neural network can be reduced. In these studied problems of tuning both network structure and parameters of a feedforward neural network, there are many parameters and numerous local optima so that these studied problems are challenging enough for evaluating the performances of any proposed GA-based approaches. The computational experiments show that the presented HTGA approach can obtain better results than the existing method reported recently in the literature.

Microstructure and Mechanical Properties of Microwave Sintered ZrO2 Bioceramics with TiO2 Addition.

The microwave sintered zirconia ceramics with 0, 1, 3, and 5 wt% TiO2 addition at a low sintering temperature of 1300°C and a short holding time of 1 hour were investigated. Effect of contents of TiO2 addition on microstructure and mechanical properties of microwave sintered zirconia bioceramics was reported. In the sintered samples, the main phase is monoclinic zirconia (m-ZrO2) phase and minor phase is tetragonal zirconia (t-ZrO2) phase. The grain sizes increased with increasing the TiO2 contents under the sintering temperature of 1300°C. Although the TiO2 phase was not detected in the XRD pattern, Ti and O elements were detected in the EDS analysis. The presence of TiO2 effectively improved grain growth of the ZrO2 ceramics. The Vickers hardness was in the range of 125 to 300 Hv and increased with the increase of TiO2 contents. Sintering temperature dependence on the Vickers hardness was also investigated from 1150°C to 1300°C, showing the increase of Vickers hardness with the increase of the sintering temperature as well as TiO2 addition.

Noninvasive evaluation of mental stress using by a refined rough set technique based on biomedical signals.

OBJECTIVE: Evaluating and treating of stress can substantially benefits to people with health problems. Currently, mental stress evaluated using medical questionnaires. However, the accuracy of this evaluation method is questionable because of variations caused by factors such as cultural differences and individual subjectivity. Measuring of biomedical signals is an effective method for estimating mental stress that enables this problem to be overcome. However, the relationship between the levels of mental stress and biomedical signals remain poorly understood. METHODS AND MATERIALS: A refined rough set algorithm is proposed to determine the relationship between mental stress and biomedical signals, this algorithm combines rough set theory with a hybrid Taguchi-genetic algorithm, called RS-HTGA. Two parameters were used for evaluating the performance of the proposed RS-HTGA method. A dataset obtained from a practice clinic comprising 362 cases (196 male, 166 female) was adopted to evaluate the performance of the proposed approach. RESULTS: The empirical results indicate that the proposed method can achieve acceptable accuracy in medical practice. Furthermore, the proposed method was successfully used to identify the relationship between mental stress levels and bio-medical signals. In addition, the comparison between the RS-HTGA and a support vector machine (SVM) method indicated that both methods yield good results. The total averages for sensitivity, specificity, and precision were greater than 96%, the results indicated that both algorithms produced highly accurate results, but a substantial difference in discrimination existed among people with Phase 0 stress. The SVM algorithm shows 89% and the RS-HTGA shows 96%. Therefore, the RS-HTGA is superior to the SVM algorithm. The kappa test results for both algorithms were greater than 0.936, indicating high accuracy and consistency. The area under receiver operating characteristic curve for both the RS-HTGA and a SVM method were greater than 0.77, indicating a good discrimination capability. CONCLUSIONS: In this study, crucial attributes in stress evaluation were successfully recognized using biomedical signals, thereby enabling the conservation of medical resources and elucidating the mapping relationship between levels of mental stress and candidate attributes. In addition, we developed a prototype system for mental stress evaluation that can be used to provide benefits in medical practice.

Fuzzy logic-based prognostic score for outcome prediction in esophageal cancer.

Given the poor prognosis of esophageal cancer and the invasiveness of combined modality treatment, improved prognostic scoring systems are needed. We developed a fuzzy logic-based system to improve the predictive performance of a risk score based on the serum concentrations of C-reactive protein (CRP) and albumin in a cohort of 271 patients with esophageal cancer before radiotherapy. Univariate and multivariate survival analyses were employed to validate the independent prognostic value of the fuzzy risk score. To further compare the predictive performance of the fuzzy risk score with other prognostic scoring systems, time-dependent receiver operating characteristic curve (ROC) analysis was used. Application of fuzzy logic to the serum values of CRP and albumin increased predictive performance for 1-year overall survival (AUC=0.773) compared with that of a single marker (AUC=0.743 and 0.700 for CRP and albumin, respectively), where the AUC denotes the area under curve. This fuzzy logic-based approach also performed consistently better than the Glasgow Prognostic Score (GPS) (AUC=0.745). Thus, application of fuzzy logic to the analysis of serum markers can more accurately predict the outcome for patients with esophageal cancer.