Ten-fold Cross-validation Experiment Research Articles

Enhancing heart failure treatment decisions: interpretable machine learning models for advanced therapy eligibility prediction using EHR data

Timely and accurate referral of end-stage heart failure patients for advanced therapies, including heart transplants and mechanical circulatory support, plays an important role in improving patient outcomes and saving costs. However, the decision-making process is complex, nuanced, and time-consuming, requiring cardiologists with specialized expertise and training in heart failure and transplantation. In this study, we propose two logistic tensor regression-based models to predict patients with heart failure warranting evaluation for advanced heart failure therapies using irregularly spaced sequential electronic health records at the population and individual levels. The clinical features were collected at the previous visit and the predictions were made at the very beginning of the subsequent visit. Patient-wise ten-fold cross-validation experiments were performed. Standard LTR achieved an average F1 score of 0.708, AUC of 0.903, and AUPRC of 0.836. Personalized LTR obtained an F1 score of 0.670, an AUC of 0.869 and an AUPRC of 0.839. The two models not only outperformed all other machine learning models to which they were compared but also improved the performance and robustness of the other models via weight transfer. The AUPRC scores of support vector machine, random forest, and Naive Bayes are improved by 8.87%, 7.24%, and 11.38%, respectively. The two models can evaluate the importance of clinical features associated with advanced therapy referral. The five most important medical codes, including chronic kidney disease, hypotension, pulmonary heart disease, mitral regurgitation, and atherosclerotic heart disease, were reviewed and validated with literature and by heart failure cardiologists. Our proposed models effectively utilize EHRs for potential advanced therapies necessity in heart failure patients while explaining the importance of comorbidities and other clinical events. The information learned from trained model training could offer further insight into risk factors contributing to the progression of heart failure at both the population and individual levels.

FreeGaze: A Framework for 3D Gaze Estimation Using Appearance Cues from a Facial Video

Gaze is a significant behavioral characteristic that can be used to reflect a person’s attention. In recent years, there has been a growing interest in estimating gaze from facial videos. However, gaze estimation remains a challenging problem due to variations in appearance and head poses. To address this, a framework for 3D gaze estimation using appearance cues is developed in this study. The framework begins with an end-to-end approach to detect facial landmarks. Subsequently, we employ a normalization method and improve the normalization method using orthogonal matrices and conduct comparative experiments to prove that the improved normalization method has a higher accuracy and a lower computational time in gaze estimation. Finally, we introduce a dual-branch convolutional neural network, named FG-Net, which processes the normalized images and extracts eye and face features through two branches. The extracted multi-features are then integrated and input into a fully connected layer to estimate the 3D gaze vectors. To evaluate the performance of our approach, we conduct ten-fold cross-validation experiments on two public datasets, namely MPIIGaze and EyeDiap, achieving remarkable accuracies of 3.11° and 2.75°, respectively. The results demonstrate the high effectiveness of our proposed framework, showcasing its state-of-the-art performance in 3D gaze estimation.

Decoding motor imagery based on dipole feature imaging and a hybrid CNN with embedded squeeze-and-excitation block

Motor imagery (MI) decoding is the core of an intelligent rehabilitation system in brain computer interface, and it has a potential advantage by using source signals, which have higher spatial resolution and the same time resolution compared to scalp electroencephalography (EEG). However, how to delve and utilize the personalized frequency characteristic of dipoles for improving decoding performance has not been paid sufficient attention. In this paper, a novel dipole feature imaging (DFI) and a hybrid convolutional neural network (HCNN) with an embedded squeeze-and-excitation block (SEB), denoted as DFI-HCNN, are proposed for decoding MI tasks. EEG source imaging technique is used for brain source estimation, and each sub-band spectrum powers of all dipoles are calculated through frequency analysis and band division. Then, the 3D space information of dipoles is retrieved, and by using azimuthal equidistant projection algorithm it is transformed to a 2D plane, which is combined with nearest neighbor interpolation to generate multi sub-band dipole feature images. Furthermore, a HCNN is designed and applied to the ensemble of sub-band dipole feature images, from which the importance of sub-bands is acquired to adjust the corresponding attentions adaptively by SEB. Ten-fold cross-validation experiments on two public datasets achieve the comparatively higher decoding accuracies of 84.23% and 92.62%, respectively. The experiment results show that DFI is an effective feature representation, and HCNN with an embedded SEB can enhance the useful frequency information of dipoles for improving MI decoding.

Multi-channel EEG-based sleep staging using brain functional connectivity and domain adaptation

Objective. Sleep stage recognition has essential clinical value for evaluating human physical/mental condition and diagnosing sleep-related diseases. To conduct a five-class (wake, N1, N2, N3 and rapid eye movement) sleep staging task, twenty subjects with recorded six-channel electroencephalography (EEG) signals from the ISRUC-SLEEP dataset is used. Approach. Unlike the exist methods ignoring the channel coupling relationship and non-stationarity characteristics, we developed a brain functional connectivity method to provide a new insight for multi-channel analysis. Furthermore, we investigated three frequency-domain features: two functional connectivity estimations, i.e. synchronization likelihood (SL) and wavelet-based correlation (WC) among four frequency bands, and energy ratio (ER) related to six frequency bands, respectively. Then, the Gaussian support vector machine (SVM) method was used to predict the five sleep stages. The performance of the applied features is evaluated in both subject dependence experiment by ten-fold cross validation and subject independence experiment by leave-one-subject-out cross-validation, respectively. Main results. In subject dependence experiment, the results showed that the fused feature (fusion of SL, WC and ER features) contributes significant gain the performance of SVM classifier, where the mean of classification accuracy can achieve 83.97% ± 1.04%. However, in subject-independence experiment, the individual differences EEG patterns across subjects leads to inferior accuracy. Five typical domain adaptation (DA) methods were applied to reduce the discrepancy of feature distributions by selecting the optimal subspace dimension. Results showed that four DA methods can significantly improve the mean accuracy by 1.89%–5.22% compared to the baseline accuracy 57.44% in leave-one-subject-out cross-validation. Significance. Compared with traditional time-frequency and nonlinear features, brain functional connectivity features can capture the correlation between different brain regions. For the individual EEG response differences, domain adaptation methods can transform features to improve the performance of sleep staging algorithms.

Cardiac disease prediction using AI algorithms with SelectKBest.

Atherosclerotic cardiovascular disease (ASCVD), which includes coronary heart disease (CHD) and ischemic stroke, is the leading cause of mortality globally. According to the European Society of Cardiology (ESC), 26 million people worldwide have heart disease, with 3.6 million diagnosed each year. Early detection of heart disease will aid in lowering the mortality rate. The lack of diversity in training data and the difficulty in comprehending the findings of complicated AI models are the key issues in current research for heart disease prediction using artificial intelligence. To overcome this, in this paper, cardiac disease prediction using AI algorithms with SelectKBest has been proposed. Features are standardized, balanced, and selected using the StandardScaler, SMOTE, and SelectKBest techniques. Machine learning models such as support vector machine (SVM), K-nearest neighbor(KNN), decision tree (DT), logistic regression (LR), adaptive boosting (AB), naive Bayes (NB), random forest (RF), and extra tree (ET) and deep learning models such as vanilla long short-term memory (LSTM), bidirectional long short-term memory (LSTM), stacked long short-term memory (LSTM), and deep neural network (DNN) are assessed using Alizadeh Sani, combined (Cleveland, Hungarian, Switzerland, Long Beach VA, and Stalog), and Pakistan heart failure datasets. As a result of the evaluation, the proposed deep neural network (DNN) with SelectKBest predicted heart disease in a promising way. The prediction rate of unweighted accuracy of 99% on Alizadeh Sani, 98% on combined, and 97% on Pakistan are gained in tenfold cross-validation experiments. The suggested approach can be utilized to diagnose heart disease in its early stages.

LDAEXC: LncRNA-Disease Associations Prediction with Deep Autoencoder and XGBoost Classifier.

Numerous scientific evidences have revealed that long non-coding RNAs (lncRNAs) are involved in the progression of human complex diseases and biological life activities. Therefore, identifying novel and potential disease-related lncRNAs is helpful to diagnosis, prognosis and therapy of many human complex diseases. Since traditional laboratory experiments are cost and time-consuming, a great quantity of computer algorithms have been proposed for predicting the relationships between lncRNAs and diseases. However, there are still much room for the improvement. In this paper, we introduce an accurate framework named LDAEXC to infer LncRNA-Disease Associations with deep autoencoder and XGBoost Classifier. LDAEXC utilizes different similarity views of lncRNAs and human diseases to construct features for each data sources. Then, the reduced features are obtained by feeding the constructed feature vectors into a deep autoencoder, and at last an XGBoost classifier is leveraged to calculate the latent lncRNA-disease-associated scores using reduced features. The fivefold cross-validation experiments on four datasets showed that LDAEXC reached AUC scores of 0.9676 ± 0.0043, 0.9449 ± 0.022, 0.9375 ± 0.0331 and 0.9556 ± 0.0134, respectively, significantly higher than other advanced similar computer methods. Extensive experiment results and case studies of two complex diseases (colon and breast cancers) further indicated the practicability and excellent prediction performance of LDAEXC in inferring unknown lncRNA-disease associations. TLDAEXC utilizes disease semantic similarity, lncRNA expression similarity, and Gaussian interaction profile kernel similarity of lncRNAs and diseases for feature construction. The constructed features are fed to a deep autoencoder to extract reduced features, and an XGBoost classifier is used to predict the lncRNA-disease associations based on the reduced features. The fivefold and tenfold cross-validation experiments on a benchmark dataset showed that LDAEXC could achieve AUC scores of 0.9676 and 0.9682, respectively, significantly higher than other state-of-the-art similar methods.

A Systematic Study of Intercultural Listening Strategies in Higher Education Institutions in the Context of Big Data

Abstract As society keeps moving forward, intercultural listening strategy is the way to promote international communication skills, and big data technology is needed to provide technical support for intercultural listening strategy in higher education institutions. In this paper, based on the logistic regression algorithm in the context of big data, an intercultural listening strategy model for higher education institutions is established, and the performance of the model is tested in a ten-fold cross-validation experiment, and four indicators are proposed in the intercultural listening strategy model to verify the effectiveness of the model. From the data, the average precision, average recall, average -value, and average -value of the ten sets of test data of the logistic regression model based on big data technology was 77.26%, 77.81%, 77.93%, and 77.56%, respectively, which indicated that all the indicators of the model met the expected performance requirements. From the four example indicators, the mean values of intercultural cognition, intercultural attitude, intercultural behavioral orientation, and listening ability were 58.2%, 60%, 60.61%, and 61.49%, respectively, and overall the higher education institutions had a clear understanding of intercultural listening strategies. The logistic regression model based on big data technology can well analyze the good and bad of current intercultural listening strategies in higher education institutions, and based on the analysis results, we can better improve the deficiencies and then promote the update of intercultural listening strategies in higher education institutions so that students in higher education institutions can show their confidence in the process of international communication and improve their intercultural communication skills.

ELUCNN for explainable COVID-19 diagnosis.

COVID-19 is a positive-sense single-stranded RNA virus caused by a strain of coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Several noteworthy variants of SARS-CoV-2 were declared by WHO as Alpha, Beta, Gamma, Delta, and Omicron. Till 13/Dec/2022, it has caused 6.65 million death tolls, and over 649 million confirmed positive cases. Based on the convolutional neural network (CNN), this study first proposes a ten-layer CNN as the backbone model. Then, the exponential linear unit (ELU) is introduced to replace ReLU, and the traditional convolutional block is now transformed into conv-ELU. Finally, an ELU-based CNN (ELUCNN) model is proposed for COVID-19 diagnosis. Besides, the MDA strategy is used to enhance the size of the training set. We develop a mobile app integrating ELUCNN, and this web app is run on a client-server modeled structure. Ten runs of the tenfold cross-validation experiment show our model yields a sensitivity of , a specificity of , an accuracy of , and an F1 score of . The ELUCNN model and mobile app are effective in COVID-19 diagnosis and give better results than 14 state-of-the-art COVID-19 diagnosis models concerning accuracy.

Deep Learning Classification of Treatment Response in Diabetic Painful Neuropathy: A Combined Machine Learning and Magnetic Resonance Neuroimaging Methodological Study

Functional magnetic resonance imaging (fMRI) has been shown successfully to assess and stratify patients with painful diabetic peripheral neuropathy (pDPN). This supports the idea of using neuroimaging as a mechanism-based technique to individualise therapy for patients with painful DPN. The aim of this study was to use deep learning to predict treatment response in patients with pDPN using resting state functional imaging (rs-fMRI). We divided 43 painful pDPN patients into responders and non-responders to lidocaine treatment (responders n = 29 and non-responders n = 14). We used rs-fMRI to extract functional connectivity features, using group independent component analysis (gICA), and performed automated treatment response deep learning classification with three-dimensional convolutional neural networks (3D-CNN). Using gICA we achieved an area under the receiver operating characteristic curve (AUC) of 96.60% and F1-Score of 95% in a ten-fold cross validation (CV) experiment using our described 3D-CNN algorithm. To our knowledge, this is the first study utilising deep learning methods to classify treatment response in pDPN.

455-P: Artificial Intelligence Approach to Treatment Classification in Painful Diabetic Neuropathy

Aim: Our study predicted treatment responses in patients with painful DPN (diabetic peripheral neuropathy) by developing a deep learning model using resting state functional magnetic resonance imaging (fMRI) neuroimaging datasets. Methods: Forty-three consecutive patients who received intravenous lidocaine treatment for painful DPN were assessed. All subjects (responders n=29 and non-responders n=14) underwent detailed clinical and neurophysiological assessments to phenotype their pain sensory profile. Subjects also underwent brain resting-state fMRI. After pre-processing we performed a group concatenated independent component analysis (ICA) set to 30 components and automatically chose 7 highly correlated (p<0.05) ICA components. A 3D convolutional neural network (CNN) classification framework was trained using a VoxNet based architecture. This deep learning architecture compared models using (1) 7 correlated ICA networks 2) all 30 ICA networks generated 3) pre-processed resting state images. Results: The deep learning treatment response classification in a ten-fold cross validation experiment using 7 ICA spatial maps has a mean AUC of 0.85 and an F1-Score of 0.90. However, with the extra information of all 30 ICA maps the mean AUROC increased to 0.97 with an F1-Score of 0.95. Using only pre-processed resting-state fMRI data achieved suboptimal F1-Score of 69% and AUC score of 44%. Conclusion: Through the use of our deep learning model, we have demonstrated high classification performance. Our method improves painful DPN treatment efficiency by stratifying patients to receive the correct treatment from the outset. We believe to our knowledge this is the first study utilising deep learning methods to classify treatment response in painful DPN. Disclosure K.Teh: None. I.D.Wilkinson: None. G.P.Sloan: None. S.Tesfaye: Advisory Panel; Astellas Pharma Inc., Bayer AG, Grünenthal Group, Nevro Corp., Wörwag Pharma GmbH & Co. KG, Speaker's Bureau; Eva Pharma, Pfizer Inc., Viatris Inc. D.Selvarajah: None. Funding National Instituteof Health Research Efficacy and Mechanism Evaluation Programme (NIHR 129921) European Foundation for the Study of Diabetes (Microvascular Complications Project Grant) University of Sheffield, Health Education England, Knowledge Exchange

Prediction of Pulmonary Function Parameters Based on a Combination Algorithm.

Objective: Pulmonary function parameters play a pivotal role in the assessment of respiratory diseases. However, the accuracy of the existing methods for the prediction of pulmonary function parameters is low. This study proposes a combination algorithm to improve the accuracy of pulmonary function parameter prediction. Methods: We first established a system to collect volumetric capnography and then processed the data with a combination algorithm to predict pulmonary function parameters. The algorithm consists of three main parts: a medical feature regression structure consisting of support vector machines (SVM) and extreme gradient boosting (XGBoost) algorithms, a sequence feature regression structure consisting of one-dimensional convolutional neural network (1D-CNN), and an error correction structure using improved K-nearest neighbor (KNN) algorithm. Results: The root mean square error (RMSE) of the pulmonary function parameters predicted by the combination algorithm was less than 0.39L and the R2 was found to be greater than 0.85 through a ten-fold cross-validation experiment. Conclusion: Compared with the existing methods for predicting pulmonary function parameters, the present algorithm can achieve a higher accuracy rate. At the same time, this algorithm uses specific processing structures for different features, and the interpretability of the algorithm is ensured while mining the feature depth information.

Spatial-Temporal Feature Fusion Neural Network for EEG-Based Emotion Recognition

The temporal and spatial information of electroencephalogram (EEG) are essential for the emotion recognition model to learn the discriminative features. Hence, we propose a novel hybrid spatial-temporal feature fusion neural network (STFFNN) to extract the discriminative features and integrate complementary information. The generated power topographic maps, which capture dependencies among the electrodes, are fed to convolutional neural network (CNN) for spatial feature learning. Furthermore, instance normalizations (INs) and batch normalizations (BNs) within the CNN are appropriately combined to alleviate the individual difference and preserve the domain-invariant information. Meanwhile, a feedforward network is adopted for temporal feature learning. Due to the high dimensionality of EEG features, we propose a grid-search-based configurational optimization method to robustly reduce the dimensionality. Finally, inspired by the multimodal fusion strategies that leverage the complementarity of data to obtain more robust predictions, we utilize a bidirectional long short-term memory (Bi-LSTM) network for temporal and spatial feature fusion. To validate the effectiveness of the proposed method, the tenfold cross-validation experiments and subject-dependent experiments are both conducted on the DEAP database. The experimental results demonstrate that the proposed method achieves outstanding performance in emotion recognition with arousal and valence level.

Estimation of drug exposure by machine learning based on simulations from published pharmacokinetic models: The example of tacrolimus

We previously demonstrated that Machine learning (ML) algorithms can accurately estimate drug area under the curve (AUC) of tacrolimus or mycophenolate mofetil (MMF) based on limited information, as well as or even better than maximum a posteriori Bayesian estimation (MAP-BE). However, the major limitation in the development of such ML algorithms is the limited availability of large databases of concentration vs. time profiles for such drugs. The objectives of this study were: (i) to develop a Xgboost model to estimate tacrolimus inter-dose AUC based on concentration-time profiles obtained from a literature population pharmacokinetic (POPPK) model using Monte Carlo simulation; and (ii) to compare its performance with that of MAP-BE in external datasets of rich concentration-time profiles. The population parameters of a previously published PK model were used in the mrgsolve R package to simulate 9000 rich interdose tacrolimus profiles (one concentration simulated every 30 min) at steady-state. Data splitting was performed to obtain a training set (75%) and a test set (25%). Xgboost algorithms able to estimate tacrolimus AUC based on 2 or 3 concentrations were developed in the training set and the model with the lowest RMSE in a ten-fold cross-validation experiment was evaluated in the test set, as well as in 4 independent, rich PK datasets from transplant patients. ML algorithms based on 2 or 3 concentrations and a few covariates yielded excellent AUC estimation in the external validation datasets (relative bias < 5% and relative RMSE < 10%), comparable to those obtained with MAP-BE. In conclusion, Xgboost machine learning models trained on concentration-time profiles simulated using literature POPPK models allow accurate tacrolimus AUC estimation based on sparse concentration data. This study paves the way to the development of artificial intelligence at the service of precision therapeutic drug monitoring in different therapeutic areas.

A cluster-based intelligence ensemble learning method for classification problems

Classification is a vital task in machine learning. By learning patterns of samples of known categories, the model can develop the ability to distinguish the categories of samples of unknown categories. Noticing the advantages of the clustering method in cluster structure analysis, we combine the clustering and classification methods to develop the novel cluster-based intelligence ensemble learning (CIEL) method. We use the clustering method to analyze the inherent distribution of the data and divide all the samples into clusters according to the characteristics of the dataset. Then, for each specific cluster, we use different classification algorithms to establish the corresponding classification model. Finally, we integrate the prediction results of each base classifier to form the final prediction result. In view of the problem of parameter sensitivity, we use a swarm intelligence algorithm to optimize the key parameters involved in the clustering, classification, and ensemble stages in order to boost the classification performance. To assess the effectiveness of CIEL, we perform tenfold cross-validation experiments on the 24 benchmark datasets provided by UCI and KEEL. Designed to improve the performance of the classifiers, CIEL outperforms other popular machine learning methods such as naive Bayes, k-nearest neighbors, random forest, and support vector machine.

Sea Ice Concentration Estimation: Using Passive Microwave and SAR Data With a U-Net and Curriculum Learning

Ice concentration estimates are typically acquired from algorithms using passive microwave satellite data, and from image analysis charts, but these have limitations. Estimates acquired from passive microwave data have coarse spatial resolution, may have errors due to atmospheric contamination, and often perform poorly in marginal ice zones. Image analysis charts are not as precise, subject to analyst interpretation, and only available over specific geographic areas. We have implemented a U-net with synthetic aperture radar images as inputs and use ice concentration estimates retrieved from passive microwave data as training labels. The U-net, due to not being sensitive to patch size, is shown to be an improvement over previous work with convolutional neural networks that use fully connected layers at the output. Data augmentation and an L1 loss function were applied along with a novel training scheme that leverages curriculum learning. In this training scheme, the model is first trained with samples from open water and consolidated ice regions before incorporating samples from marginal ice regions. In a tenfold cross validation experiment, we achieve 3–4% mean absolute error comparing to estimates using passive microwave data and observe curriculum learning models having more stable training. Predictions on four with-held SAR scenes with difficult ice conditions were evaluated with image analysis charts. A mean absolute error of 7.18% is achieved, which is lower than errors associated with passive microwave data alone. Qualitative improvements in marginal ice zone estimates are achieved, while still preserving smooth consolidated ice regions, and openings in ice cover.

Inferring Potential CircRNA-Disease Associations via Deep Autoencoder-Based Classification.

Circular RNAs (circRNA) are endogenous non-coding RNA molecules with a stable circular conformation. Growing evidence from recent experiments reveals that dysregulations and abnormal expressions of circRNAs are correlated with complex diseases. Therefore, identifying the causal circRNAs behind diseases is invaluable in explaining the disease pathogenesis. Since biological experiments are difficult, slow-progressing, and prohibitively expensive, computational approaches are necessary for identifying the relationships between circRNAs and diseases. We propose an ensemble method called AE-RF, based on a deep autoencoder and random forest classifier, to predict potential circRNA-disease associations. The method first integrates circRNA and disease similarities to construct features. The integrated features are sent to the deep autoencoder, to extract hidden biological patterns. With the extracted deep features, the random forest classifier is trained for association prediction. AE-RF achieved AUC scores of 0.9486 and 0.9522, in fivefold and tenfold cross-validation experiments, respectively. We conducted case studies on the top-most predicted results and three common human cancers. We compared the method with state-of-the-art classifiers and related methods. The experimental results and case studies demonstrate the prediction power of the model, and it outperforms previous methods with high degree ofrobustness. Training the classifier with the unique features retrieved by the autoencoder enhanced the model's predictive performance. The top predicted circRNAs are promising candidates for further biological tests.

A Machine Learning Approach to Estimate the Glomerular Filtration Rate in Intensive Care Unit Patients Based on Plasma Iohexol Concentrations and Covariates.

This work aims to evaluate whether a machine learning approach is appropriate to estimate the glomerular filtration rate in intensive care unit patients based on sparse iohexol pharmacokinetic data and a limited number of predictors. Eighty-six unstable patients received 3250mg of iohexol intravenously and had nine blood samples collected 5, 30, 60, 180, 360, 540, 720, 1080, and 1440min thereafter. Data splitting was performed to obtain a training (75%) and a test set (25%). To estimate the glomerular filtration rate, 37 candidate potential predictors were considered and the best machine learning approach among multivariate-adaptive regression spline and extreme gradient boosting (Xgboost) was selected based on the root-mean-square error. The approach associated with the best results in a ten-fold cross-validation experiment was then used to select the best limited combination of predictors in the training set, which was finally evaluated in the test set. The Xgboost approach yielded the best performance in the training set. The best combination of covariates comprised iohexol concentrations at times 180 and 720min; the relative deviation from these theoretical times; the difference between these two concentrations; the Simplified Acute Physiology Score II; serum creatinine; and the fluid balance. It resulted in a root-mean-square error of 6.2mL/min and an r2 of 0.866 in the test set. Interestingly, the eight patients in the test set with a glomerular filtration rate < 30mL/min were all predicted accordingly. Xgboost provided accurate glomerular filtration rate estimation in intensive care unit patients based on two timed blood concentrations after iohexol intravenous administration and three additional predictors.

Breathing Sound Segmentation and Detection Using Transfer Learning Techniques on an Attention-Based Encoder-Decoder Architecture.

This paper focuses on the use of an attention-based encoder-decoder model for the task of breathing sound segmentation and detection. This study aims to accurately segment the inspiration and expiration of patients with pulmonary diseases using the proposed model. Spectrograms of the lung sound signals and labels for every time segment were used to train the model. The model would first encode the spectrogram and then detect inspiratory or expiratory sounds using the encoded image on an attention-based decoder. Physicians would be able to make a more precise diagnosis based on the more interpretable outputs with the assistance of the attention mechanism.The respiratory sounds used for training and testing were recorded from 22 participants using digital stethoscopes or anti-noising microphone sets. Experimental results showed a high 92.006% accuracy when applied 0.5 second time segments and ResNet101 as encoder. Consistent performance of the proposed method can be observed from ten-fold cross-validation experiments.

Emergency Department Capacity Planning: A Recurrent Neural Network and Simulation Approach.

Emergency departments (EDs) play a vital role in the whole healthcare system as they are the first point of care in hospitals for urgent and critically ill patients. Therefore, effective management of hospital's ED is crucial in improving the quality of the healthcare service. The effectiveness depends on how efficiently the hospital resources are used, particularly under budget constraints. Simulation modeling is one of the best methods to optimize resources and needs inputs such as patients' arrival time, patient's length of stay (LOS), and the route of patients in the ED. This study develops a simulation model to determine the optimum number of beds in an ED by minimizing the patients' LOS. The hospital data are analyzed, and patients' LOS and the route of patients in the ED are determined. To determine patients' arrival times, the features associated with patients' arrivals at ED are identified. Mean arrival rate is used as a feature in addition to climatic and temporal variables. The exhaustive feature-selection method has been used to determine the best subset of the features, and the mean arrival rate is determined as one of the most significant features. This study is executed using the one-year ED arrival data together with five-year (43.824 study hours) ED arrival data to improve the accuracy of predictions. Furthermore, ten different machine learning (ML) algorithms are used utilizing the same best subset of these features. After a tenfold cross-validation experiment, based on mean absolute percentage error (MAPE), the stateful long short-term memory (LSTM) model performed better than other models with an accuracy of 47%, followed by the decision tree and random forest methods. Using the simulation method, the LOS has been minimized by 7% and the number of beds at the ED has been optimized.

Structural risk minimization of rough set-based classifier

The classification ability in unseen objects, namely generalization ability, remains a long-standing challenge in rough set-based classifier. Current research mainly focuses on introducing thresholds to tolerate some errors in seen objects. The reason for introducing thresholds and the selection of threshold still lack sufficient theoretical support. The structural risk minimization (SRM) inductive principle is one of the most effective theories to control the generalization ability, which suggests a trade-off between errors in seen objects and complexity. Therefore, this paper introduces the SRM principle into rough set-based classifier and proposes SRM algorithm of rough set-based classifier called SRM-R algorithm. SRM-R algorithm uses the number of rules to characterize the actual complexity of rough set-based classifier and obtains the optimal trade-off between errors in seen objects and complexity through genetic multi-objective optimization. The tenfold cross-validation experiment in 12 UCI datasets shows SRM-R algorithm can significantly improve the generalization ability compared with conventional threshold algorithm. Besides, this paper uses other two possible complexity metrics including the number of attributes and attribute space to construct corresponding SRM algorithms, respectively, and compared their classification accuracy with that of SRM-R algorithm. Comparison result shows SRM-R algorithm obtains optimal classification accuracy. This indicates that the number of rules characterizes the complexity more effectively than the number of attributes and attribute space. Further experiments show that SRM-R algorithm obtains fewer rules and larger support coefficient, which means it extracts stronger rules. This explains why it obtains better generalization ability to some extent.