Test Data Points Research Articles

NPM1 Minimal Residual Disease Detection at Any Time by Clinical Next Generation Sequencing Predicts Poor Survival in Acute Myeloid Leukemia

Abstract Introduction/Objective Acute Myeloid Leukemia (AML) is a deadly disease that accounts for 80% of adult leukemia cases and carries a 5-year survival rate of only 31.7%. Historically, prognosis hinged heavily on cytogenetic stratification, but more recently minimal residual disease (MRD) has emerged as a crucial prognostic marker. Flow cytometry, PCR, and NGS can all be employed for MRD detection, but each carries its own limitations. Though technically challenging, NGS has the advantage being able to simultaneously assay multiple mutations, thus providing for broader applicability. NPM1 is the most prevalently mutated gene in AML (25-35%). In this study, one of the largest of its kind, we investigate the feasibility and impact of NPM1 MRD detection by clinical NGS. Methods/Case Report IRB approval was obtained. We reviewed the medical record and clinical laboratory data sets from AML patients treated at Moffitt Cancer Center who had NPM1 status assessed by clinical NGS, Testing was performed at LabPMM (San Diego, CA). DNA was extracted, and NPM1 exon 12 was amplified. PCR products were sequenced on an Illumina MiSeq. The clinically reportable assay sensitivity was 5 x 10-5 though lower levels were detected on review of raw data. Kaplan-Meier analysis was conducted for overall survival assessment. Since patients often had multiple tests, we divide patients into those with NPM1+ at any time and NPM1 always negative. Results (if a Case Study enter NA) There was a total of 140 patients. Overall, 346 tests were conducted, with an average of 2.5 tests/patient. There were 85 patients (60%) who underwent serial testing. MRD+ patients (n=80) had a mean age of 66 years, while MRD- patients (n=60) had a mean age of 58 years. MRD+ patients exhibited a mean overall survival of 29.3 months versus 43.6 months for MRD- patients (p=0.005.) Conclusion Through our study encompassing a large patient cohort of 140 patients and nearly 350 testing data points, we confirm the feasibility and clinical utility of NPM1 MRD evaluation by NGS. The significantly poorer survival in patients who were MRD+ at any time during their disease provides a scientific rationale for serial NGS testing to enable risk stratification of NPM1+ AML patients.

Classification of Simulated Fake Bandwidth Data Using LSTM

Hardly will someone acknowledge that the bandwidth we use every day is as authentic as most ISPs advertise, even those offering dedicated services. There are usually shortcomings, especially on upload and download bandwidth speeds. This paper presents the classification of simulated fake bandwidth data using the Long Short-Term Memory model, which though seldom found, is a very effective approach in network analysis. There were 1400 bandwidth data points collected from the MikroTik RB 1100 AHx device in a month, then further processed with normalization, and divided to have 80% training and 20% testing. The LSTM model applied had an accuracy rate of 98.93%, proving that it is capable of classifying either genuine or fake bandwidth instances accordingly. Of 1,400 test data points, the model managed to classify 723 as fake bandwidth and another 677 as genuine, resulting in a classification error rate of only 1.07%. The results clearly prove that LSTM has huge potential for real-time bandwidth manipulation detection, key to enhancing trust and efficiency in network management. In this respect, this research shows that bandwidth analysis combined with LSTM can be an original solution for network monitoring

OPTIMIZING AN EXPERT SYSTEM FOR DIAGNOSING A DEPRESSION DISORDER USING A CASE BASED REASONING METHOD

According to data from the World Health Organization (WHO), 3.7% of the population in Indonesia experiences depression. Depression can impact both the mental and physical conditions of an individual. WHO reports that every year, approximately 800,000 people die by suicide, with depression being one of the causes. Depression treatment is handled by a professional, and in the field, the process of diagnosing depression disorders is still generally done manually by them. This creates many opportunities for errors, despite the fact that each level of depression disorder requires different handling. Inadequate treatment can hinder the patient's recovery and may potentially worsen their condition. A precise and efficient method is needed to diagnose depression disorders. An expert system can reduce the risk of errors that occur with manual calculations. The implemented case-based reasoning method can classify depression disorders. Testing was conducted using 30 datasets as initial knowledge, with 20 sample data points for testing, randomly selected from the population through questionnaires. The classification accuracy for depression disorders reached up to 90%.

Prediksi Kelulusan Mata Kuliah Mahasiswa Teknologi Informasi Menggunakan Algoritma K-Nearest Neighbor

This research aims to develop a predictive model using the K-Nearest Neighbor (KNN) method to forecast the course completion of students in the Information Technology program. The issue at hand is the uncertainty in predicting student success based on historical data and specific attributes. This study focuses on the importance of understanding the factors that influence student success in the Database Management Systems course to provide accurate predictions and help improve student pass rates in this course. The objective of this research is to build a predictive model using the KNN algorithm and to implement this model using the PHP programming language. The study aims to offer valuable insights for educational institutions to enhance teaching and learning processes and to expand understanding of data mining concepts in specific case studies. The prediction aims to determine whether a student will pass or fail the Database Management Systems course based on predetermined training and testing data. By calculating the nearest distance between the training data and the test data. The results showed an accuracy rate of 90% for predicting course completion using k=5, with a dataset consisting of 40 training data points and 20 testing data points.

Fluid viscosity prediction leveraging computer vision and robot interaction

Accurately determining fluid viscosity is crucial for various industrial and scientific applications. Traditional methods of viscosity measurement, though reliable, often require manual intervention and cannot easily adapt to real-time monitoring. With advancements in machine learning and computer vision, this work explores the feasibility of predicting fluid viscosity by analyzing fluid oscillations captured in video data. The pipeline employs a 3D convolutional autoencoder pretrained in a self-supervised manner to extract and learn features from semantic segmentation masks of oscillating fluids. Then, the latent representations of the input data, produced from the pretrained autoencoder, are processed with a distinct inference head to infer either the fluid category (classification) or the fluid viscosity (regression) in a time-resolved manner. When the latent representations generated by the pre-trained autoencoder are used for classification, the system achieves a 97.1% accuracy across a total of 4140 test datapoints. Similarly, for regression tasks, employing an additional fully-connected network as a regression head allows the pipeline to achieve a mean absolute error of 0.258 cP over 4416 test datapoints. This study represents an innovative contribution to both fluid characterization and the evolving landscape of Artificial Intelligence, demonstrating the potential of deep learning in achieving near real-time viscosity estimation and addressing practical challenges in fluid dynamics through the analysis of video data capturing oscillating fluid dynamics.

A Novel Movable Mannequin Platform for Evaluating and Optimising mmWave Radar Sensor for Indoor Crowd Evacuation Monitoring Applications

Developing mmWave radar sensors for indoor crowd motion sensing and tracking faces a critical challenge: the scarcity of large-scale, high-quality training data. Traditional human experiments encounter logistical complexities, ethical considerations, and safety issues. Replicating precise human movements across trials introduces noise and inconsistency into the data. To address this, this study proposes a novel solution: a movable platform equipped with a life-size mannequin to generate realistic and diverse data points for mmWave radar training and testing. Unlike human subjects, the platform allows precise control over movements, optimising sensor placement relative to the target object. Preliminary optimisation results reveal that sensor height impacts tracking performance, with an optimal sensor placement above the test subject yields the best results. The results also reveal that the 3D data format outperforms 2D data in accuracy despite having fewer frames. Additionally, analysing height distribution using 3D data highlights the importance of the sensor angle—15° downwards from the horizontal plane.

Prediction of CHF location through applied machine learning

The prediction of critical heat flux (CHF) location is a challenging issue in power generation industries. Several factors viz. geometric parameters and operating conditions influence the occurrence of CHF, and many mechanisms have been proposed to better explain the physics behind the occurrence of CHF. As CHF is a complex phenomenon, its dependency on these factors exhibit non-linear relationship. Further, it is challenging to account for all the factors simultaneously in a single model. Moreover, the benchmark experimental data has to be available to validate the constructed model with all the factors viz., operating pressure and power, inlet mass flux and subcooling etc. Due to these reasons, the influence of these factors remain unresolved. Considering the above issues, the present study employs different machine learning models to predict CHF location and identify the accurate model based on the comparisons against the chosen benchmark experimental data. Becker et al. (1983) data bank was selected for machine learning model training as well as the testing due to the wide range of operating conditions covered. Hyperparameter tuning is applied to optimize the performance of the models. Among all the implemented models, the artificial neural network (ANN) model is found to perform well with the training and the test datasets. The mean absolute percentage error (MAPE) of the training and testing data points are 4.01% and 6.04%, respectively. The accuracy score of the ANN model is approximately 96% for the training and 94% for the testing datasets. The proposed model is helpful for design engineers and analysts of power plants for design optimization and to reduce the computational time.

Forecast of solar cycle 25 based on Hybrid CNN-Bidirectional-GRU (CNN-BiGRU) model and Novel Gradient Residual Correction (GRC) technique

In this study, a hybrid deep-learning(DL) model is proposed, which consists of a Convolution-Neural-Network (CNN) and Bidirectional-Gated-Recurrent-Unit (Bi-GRU) for the prediction of sunspot numbers(SSN) of different frequencies along with novel post-processing techniques of Gradient-Residual-Correction (GRC) to enhance the accuracy of the predictions further. GRC is utilized to reduce the residual present in the forecasted values. The AdaBoost regression model is implemented over the residual obtained from the prediction of training data using a hybrid CNN-BiGRU model with respect to the gradient of the training data points to predict the residual for the test data points. Ultimately, the predicted residual for test data points is summed up with the predicted test data points to achieve the final predictions. The results obtained from the proposed methods are compared with the results obtained from traditional DL models. The validation of the proposed method is carried out based on four performance metrics, namely Root Mean Squared Error(RMSE), Mean Absolute Scaled Error(MASE), Mean Absolute Error(MAE), and Mean Absolute Percentage Error(MAPE). A significant percentage of improvement is observed while using the GRC technique in comparison to all other experimented models for all four variants of SSN data. A maximum improvement of 85.71% has been achieved in comparison to the BiLSTM model over the 13-month smoothed SSN dataset on the basis of MAPE. Friedman Ranking is also performed as a non-parametric statistical test over the results of the performance measures. This model has been utilized for the forecast of solar cycle 25(SC25) over the annual mean of total SSN. It has been observed that the SC25 is expected to reach its peak in the year 2024 with an annual average peak value of 143.641. Comparative analysis of SC25 and the peak of SSN in the ongoing cycle with the previous works are also carried out.

Performance Evaluation of Artificial Neural Network Modelling to a Ploughing Unit in Various Soil Conditions

Abstract The specific objective of this study is to find a suitable artificial neural network model for estimating the operation indicators (disturbed soil volume, effective field capacity, draft force, and energy requirement) of ploughing units (tractor disc) in various soil conditions. The experiment involved two different factors, i.e., (Ι) soil texture index and (ΙΙ) field work index, and included soil moisture content, tractor engine power, soil bulk density, tillage speed, tillage depth, and tillage width, which were linked to one dimensionless index. We assessed the effectiveness of artificial neural network and multiple linear regression models between the values predicted and the actual values using the mean absolute error criterion to test data points. When the artificial neural network model was applied, the mean absolute error values for disturbed soil volume, effective field capacity, draft force, and energy requirement were 69.41 m3·hr−1, 0.04 ha·hr−1, 1.24 kN, and 1.95 kw·hr·ha−1, respectively. In order to evaluate the behaviour of new models, the coefficient R 2 was used as a criterion, where R 2 values in artificial neural network were 0.9872, 0.9553, 0.9948, and 0.9718, respectively, for the aforementioned testing dataset. Simultaneously, R 2 values in multiple linear regression were 0.7623, 0.696, 0.492, and 0.5572, respectively, for the same testing dataset. Based on these comparisons, it was clear that predictions using the artificial neural network models proposed are very satisfactory.

Dapatkah Good Corporate Governance Mencegah Penghindaran Pajak?

ABSTRACT
 This study was conducted to examine the influence of audit quality, audit committee, independent commisioners, and institutional ownership on tax avoidance. A quantitive research method was employed. The research population consisted of 238 data points. The sampling technique used was purposive sampling, resulting in 188 data points that met the criteria. Outliers were identified, and 27 data points were removed, leaving 161 data points for testing. Data analysis techniques used in this research included descriptive statistics, classical assumption test, multiple linear regression analysis, model testing, and hypothesis testing. The result of the study that audit quality has a significant positive impact on tax avoidance, independent commissioners have a significant negative impact on tax avoidance, while the audit committee and institutional ownership do not have an impact on tax avoidance.
 Keywords : Audit Quality, Audit Committee, Independent Commisioners, Institutional Ownership, Tax Avoidance

Multivariate sample similarity measure for feature selection with a resemblance model

Feature selection improves the classification performance of machine learning models. It also identifies the important features and eliminates those with little significance. Furthermore, feature selection reduces the dimensionality of training and testing data points. This study proposes a feature selection method that uses a multivariate sample similarity measure. The method selects features with significant contributions using a machine-learning model. The multivariate sample similarity measure is evaluated using the University of California, Irvine heart disease dataset and compared with existing feature selection methods. The multivariate sample similarity measure is evaluated with metrics such as minimum subset selected, accuracy, F1-score, and area under the curve (AUC). The results show that the proposed method is able to diagnose chest pain, thallium scan, and major vessels scanned using X-rays with a high capability to distinguish between healthy and heart disease patients with a 99.6% accuracy.

Determining the adjusting bias in reactor pressure vessel embrittlement trend curve using Bayesian multilevel modelling

A sophisticated Bayesian multilevel model for estimating group bias was developed to improve the utility of the ASTM E900-15 embrittlement trend curve (ETC) to assess the conditions of nuclear power plants (NPPs). For multilevel model development, the Baseline 22 surveillance dataset was basically classified into groups based on the NPP name, product form, and notch orientation. By including the notch direction in the grouping criteria, the developed model could account for TTS differences among NPP groups with different notch orientations, which have not been considered in previous ETCs. The parameters of the multilevel model and biases of the NPP groups were calculated using the Markov Chain Monte Carlo method. As the number of data points within a group increased, the group bias approached the mean residual, resulting in reduced credible intervals of the mean, and vice versa. Even when the number of surveillance test data points was less than three, the multilevel model could estimate appropriate biases without overfitting. The model also allowed for a quantitative estimate of the changes in the bias and prediction interval that occurred as a result of adding more surveillance test data. The biases estimated through the multilevel model significantly improved the performance of E900-15.

Illicit Activity Detection in Bitcoin Transactions using Timeseries Analysis

A key motivator for the usage of cryptocurrency such as bitcoin in illicit activity is the degree of anonymity provided by the alphanumeric addresses used in transactions. This however does not mean that anonymity is built into the system as the transactions being made are still subject to the human element. Additionally, there is around 400 Gigabytes of raw data available in the bitcoin blockchain, making it a big data problem. HPCC Systems is used in this research, which is a data intensive, open source, big data platform. This paper attempts to use timing data produced by taking the time intervals between consecutive transactions performed by an address and make an identification of the nature of the address (illegal or legal). With the use of three different goodness of fit run tests namely Kolmogorov–Smirnov test, Anderson-Darling test and Cramér–von Mises criterion, two addresses are compared to find if they are from the same source. The BABD-13 dataset was used as a source of illegal addresses, which provided both references and test data points. The research shows that time-series data can be used to represent transactional behaviour of a user and the algorithm proposed is able to identify different addresses originating from the same user or users engaging in similar activity.

Artificial neural network modeling on the polymer-electrolyte aqueous two-phase systems involving biomolecules

Polymer-electrolyte aqueous two-phase systems (ATPS) have demonstrated their superior performance in the separation and purification of high-value biomolecules. However, these powerful platforms are still a major academic curiosity, without their acceptance and implementation by industry. One of the major obstacles is the absence of models to predict the partition of biomolecules in ATPS in an easy and predictive way. To address this limitation, modelling studies on the binodal curve behavior of polymer-electrolyte ATPS and the partitioning of biomolecules in these aqueous electrolyte solutions are carried out in this work. First, a comprehensive database targeting the studied systems is established. In total, 11,998 experimental binodal data points covering 276 polymer-electrolyte ATPS at different temperatures (273.15 K-399.15 K) and 626 experimental partition data points involving 22 biomolecules in 42 polymer-electrolyte ATPS at different temperatures (283.15 K-333.15 K) are included. Then, a novel modeling strategy that combines a well-known machine learning algorithm, i.e., artificial neural network (ANN) and group contribution (GC) method is proposed. Based on this modeling strategy, an ANN-GC model (ANN-GC model1) is built to describe the binodal curve behavior of polymer-electrolyte ATPS, while another ANN-GC model (ANN-GC model2) is developed to predict the partition of biomolecules in these biphasic systems. ANN-GC model1 gives a mean absolute error (MAE) of 0.0132 and squared correlation coefficient (R2) of 0.9878 for the 9,598 training data points, and for the 1,200 validation data points they are 0.0141 and 0.9858, respectively. Meanwhile, it also gives a MAE of 0.0143 and R2 of 0.9846 for the 1,200 test data points. On the other hand, ANN-GC model2 gives root-mean-square deviation (RMSD) of 0.0577 for 501 training data points, and for the 62 validation data points and 63 test data points their RMSD are 0.0849 and 0.0885, respectively. Furthermore, the obtained results also indicate that the tie-line length of polymer-electrolyte ATPS calculated from ANN-GC model1 can be directly used in ANN-GC model2 for predicting the partition performance coefficient of biomolecules in these ATPS. The developed models offer the possibility to predict the partition of biomolecules in ATPS without any requirement of experimental data. Based on the developed ANN-GC models, some high-performance ATPS are identified to partition four well-known biomolecules.

Estimation of Cerebral Blood Flow and Arterial Transit Time From Multi-Delay Arterial Spin Labeling MRI Using a Simulation-Based Supervised Deep Neural Network.

An inherently poor signal-to-noise ratio (SNR) causes inaccuracy and less precision in cerebral blood flow (CBF) and arterial transit time (ATT) when using arterial spin labeling (ASL). Deep neural network (DNN)-based parameter estimation can solve these problems. To reduce the effects of Rician noise on ASL parameter estimation and compute unbiased CBF and ATT using simulation-based supervised DNNs. Retrospective. One million simulation test data points, 17 healthy volunteers (five women and 12 men, 33.2 ± 14.6 years of age), and one patient with moyamoya disease. 3.0 T/Hadamard-encoded pseudo-continuous ASL with a three-dimensional fast spin-echo stack of spirals. Performances of DNN and conventional methods were compared. For test data, the normalized mean absolute error (NMAE) and normalized root mean squared error (NRMSE) between the ground truth and predicted values were evaluated. For in vivo data, baseline CBF and ATT and their relative changes with respect to SNR using artificial noise-added images were assessed. One-way analysis of variance with post-hoc Tukey's multiple comparison test, paired t-test, and the Bland-Altman graphical analysis. Statistical significance was defined as P < 0.05. For both CBF and ATT, NMAE and NRMSE were lower with DNN than with the conventional method. The baseline values were significantly smaller with DNN than with the conventional method (CBF in gray matter, 66 ± 10 vs. 71 ± 12 mL/100 g/min; white matter, 45 ± 6 vs. 46 ± 7mL/100 g/min; ATT in gray matter, 1424 ± 201 vs. 1471 ± 154 msec). CBF and ATT increased with decreasing SNR; however, their change rates were smaller with DNN than were those with the conventional method. Higher CBF in the prolonged ATT region and clearer contrast in ATT were identified by DNN in a clinical case. DNN outperformed the conventional method in terms of accuracy, precision, and noise immunity. 3 Technical Efficacy: Stage 1.

Artificial Neural Network Model to Predict Production Rate of Electrical Submersible Pump Wells

Summary Production data are essential for designing and operating electrical submersible pump (ESP) systems. This study aims to develop artificial neural network (ANN) models to predict flow rates of ESP artificially lifted wells. The ANN models were developed using 31,652 data points randomly split into 80% (25,744 data points) for training and 20% (5,625 data points) for testing. Each data set included measurements for wellhead parameters, fluid properties, ESP downhole sensor measurements, and variable speed drive (VSD) sensors parameters. The models consisted of four separate neural networks to predict oil, water, gas, and liquid flow rates. Sensitivity analyses were performed to determine the optimum number of input parameters that can be used in the model. The best performance was achieved with ANN models of 16 input parameters that are readily available in ESP wells. The results of the best ANN configuration indicate that the mean absolute percent error (MAPE) between the predicted flow rates and the actual measurements for the testing data points of the oil, water, gas, and liquid networks is 3.7, 5.2, 6.4, and 4.1%, respectively. In addition, the correlation coefficients (R2) are 0.991, 0.992, 0.983, and 0.979 for the estimated oil, water, gas, and liquid flow rates for the testing data points, respectively. The performance of the ANN models was compared against performance of published physics-based models and the results were comparable. Unlike the physics-based methods, the ANN models have the advantage that they do not require periodic calibration. The ANN models were used to predict the flow rate curves of an oilwell in the Western Desert of Egypt. The results were compared to the actual separator test data. It was clear that the model results matched the actual test data. The ANN model is useful for predicting individual well production rates within wide variety of pumping conditions and completion configurations. This should allow for continuous monitoring, optimization, and performance analysis of ESP wells as well as quicker response to operational issues. In comparison to traditional separators and multiphase flowmeters (MPFMs), the use of the developed ANN models is simple, quick, and inexpensive.

Prediction and uncertainty quantification of ultimate bond strength between UHPC and reinforcing steel bar using a hybrid machine learning approach

The composite action of the reinforcing bars in the UHPC involves complex and nonlinear mechanisms. Inadequate knowledge of their interaction may lead to insufficient bond resistance in the concrete structures. The behavior of reinforcing steel bars in UHPC is generally determined by conducting pull-out tests. However, these tests need amounts of time and cost. A soft computing technique is an alternative approach that exterminates the need to conduct pull-out tests. This study presents the application of artificial intelligence technique in predicting ultimate bond strength (UBS) between ultrahigh-performance concrete (UHPC) and steel reinforcement bars and performs uncertainty quantification of such values. A total of 333 pull-out test data points were collected. A hybrid machine learning (ML) model, improved eliminate particle swamp optimization hybridized artificial neural network (IEPANN) was developed to predict the UBS between UHPC and reinforcing bars. The efficacy of the IEPANN was evaluated against other ML models such as particle swamp optimization hybridized artificial neural network (PANN), support vector regression (SVR), and multiple linear regression (MLR). The IEPANN model was benchmarked against three design codes, including CEB-FIP, AS3600, EC2, and four empirical models. The result obtained based on R = 0.973, MAE = 1.880, RMSE = 2.595, MAPE = 7.708, and RI = 0.944 shows that the IEPANN model can predict UBS between UHPC and reinforcing steel bars with acceptable accuracy. The uncertainty of the UBS was quantified with geometrical and material configuration randomness based on Monte Carlo simulation. The simulation result shows that the yield strength of steel rebar is the most sensitive, while the embedment length is the least sensitive to the bond strength. Systematic assessment of uncertainties in UBS in the design of reinforced concrete members may lead to a higher level of confidence and enhance construction reliability.

Prediction of Mechanical Behaviours of FRP-Confined Circular Concrete Columns Using Artificial Neural Network and Support Vector Regression: Modelling and Performance Evaluation.

The prediction and control of the mechanical behaviours of fibre-reinforced polymer (FRP)-confined circular concrete columns subjected to axial loading are directly related to the safety of the structures. One challenge in building a mechanical model is understanding the complex relationship between the main parameters affecting the phenomenon. Artificial intelligence (AI) algorithms can overcome this challenge. In this study, 298 test data points were considered for FRP-confined circular concrete columns. Six parameters, such as the diameter-to-fibre thickness ratio (D/t) and the tensile strength of the FRP (ffrp) were set as the input sets. The existing models were compared with the test data. In addition, artificial neural networks (ANNs) and support vector regression (SVR) were used to predict the mechanical behaviour of FRP-confined circular concrete columns. The study showed that the predictive accuracy of the compressive strength in the existing models was higher than the peak compressive strain for the high dispersion of material deformation. The predictive accuracy of the ANN and SVR was higher than that of the existing models. The ANN and SVR can predict the compressive strength and peak compressive strain of FRP-confined circular concrete columns and can be used to predict the mechanical behaviour of FRP-confined circular concrete columns.

Accurate automatic classification system for 3D CT images of some vertebrate remains from Egypt

Vertebrate fossils/remains became recently significant in various study fields for determining the ecological biodiversity. However, with the great abundance of fossils/remains and their classes, there is a difficulty in identifying and detecting these classes. Hence, in this paper, an accurate machine learning classification technique is presented to differentiate automatically some types of 3D vertebrate remain images. A computed tomography (CT) scanner is utilized to construct a dataset of 3D images of some vertebrate remains found in Egypt. Adaptive enhancement and segmentation processes are applied to the dataset. The different selected geometric features are then extracted. Thus, the extracted features are classified using suitable machine learning classifiers (SVM, KNN, DTs). The automatic detection for the remains class, according to the extracted features, is obtained using the confusion matrix for the training and testing data points and the receiver operating characteristic (ROC) curve. The results confirmed an accurate technique with high performance.

Electronic Health Record-Based Deep Learning Prediction of Death or Severe Decompensation in Heart Failure Patients

BackgroundSurgical mechanical ventricular assistance and cardiac replacement therapies, although life-saving in many heart failure (HF) patients, remain high-risk. Despite this, the difficulty in timely identification of medical therapy nonresponders and the dire consequences of nonresponse have fueled early, less selective surgical referral. Patients who would have ultimately responded to medical therapy are therefore subjected to the risk and life disruption of surgical therapy. ObjectivesThe purpose of this study was to develop deep learning models based upon commonly-available electronic health record (EHR) variables to assist clinicians in the timely and accurate identification of HF medical therapy nonresponders. MethodsThe study cohort consisted of all patients (age 18 to 90 years) admitted to a single tertiary care institution from January 2009 through December 2018, with International Classification of Disease HF diagnostic coding. Ensemble deep learning models employing time-series and densely-connected networks were developed from standard EHR data. The positive class included all observations resulting in severe progression (death from any cause or referral for HF surgical intervention) within 1 year. ResultsA total of 79,850 distinct admissions from 52,265 HF patients met observation criteria and contributed >350 million EHR datapoints for model training, validation, and testing. A total of 20% of model observations fit positive class criteria. The model C-statistic was 0.91. ConclusionsThe demonstrated accuracy of EHR-based deep learning model prediction of 1-year all-cause death or referral for HF surgical therapy supports clinical relevance. EHR-based deep learning models have considerable potential to assist HF clinicians in improving the application of advanced HF surgical therapy in medical therapy nonresponders.