Predictive Regression Research Articles

Machine learning versus logistic regression for permanent pacemaker implantation prediction after transcatheter aortic valve replacement—a systematic review and meta-analysis

Machine learning versus logistic regression for permanent pacemaker implantation prediction after transcatheter aortic valve replacement—a systematic review and meta-analysis

Experimental study of NiTi alloy cardiovascular stent formed via SLM

Selective laser melting (SLM) has garnered significant attention in the manufacturing of cardiovascular stents due to its potential for fabricating customized stents with intricate geometries that meet clinical requirements. This study aims to optimize the SLM parameters for NiTi stents to achieve minimal porosity and surface roughness, and investigate the biocompatibility and compression tests of SLMed stents. Firstly, the single-factor experiments of laser power, scanning speed and hatch spacing are conducted, respectively. Subsequently, a Box-Benhnken experimental was employed to establish regression prediction models for porosity and surface roughness using response surface methodology (RSM). The results show that scanning speed is the most influential factor on porosity, while laser power significantly affects surface roughness. The optimized parameter combination with minimum porosity of 0.165% and surface roughness of 6.17 μm as follows: laser power of 156W, scanning speed of 712mm/s, and hatch spacing of 0.9mm. The relative errors of them are 0.19% and 0.8%, respectively, indicating that the regression model possesses good prediction ability. Additionally, the corrosion rate and nickel ion release gradually increase with the immersion time, reaching the highest corrosion rate of 2.46g/m² and nickel ion release of 1.08μg after 42 days. Besides, the SLMed stent demonstrated good recovery ability in the compression test. The findings promote the utilization of SLM in the manufacturing of NiTi cardiovascular stents for medical implant applications.

A probabilistic integration of LSTM and Gaussian process regression for uncertainty-aware reservoir water level predictions

ABSTRACT Reservoir-level forecasting, while being crucial for optimal operation, is challenged by complex physical processes and changing climate conditions. Machine learning approaches offer deterministic predictions but often neglect system physics and uncertainty. This article presents a probabilistic data-driven approach combining long short-term memory (LSTM) and Gaussian process regression (GPR) to provide both point forecasts and uncertainty estimates. The hybrid model leverages LSTM’s fitting capabilities with GPR’s robust Bayesian frameworks for uncertainty estimation in non-linear problems, offering accurate predictions without extensive high-fidelity modelling, and avoiding frequent training and parameter optimization. Evaluation with real reservoir data from India shows the model’s superiority over the vanilla LSTM for both univariate and multivariate scenarios. The proposed model achieved a Nash-Sutcliffe efficiency of 0.97 to 0.98, a mean biased error of −0.5634 to −1.0314 for 10-day forecasts, and a continuous ranked probability score of 5.80 and 1.87 for the Bhakra and Pong reservoirs, respectively.

Predictive Modeling in IoT-Driven Agriculture: A Comparative Study of Regression and Classification with LIME Interpretability

Abstract. This study aims to compare the performance of regression and classification machine learning models in crop prediction by analyzing data from IoT-driven agriculture. Linear regression and random forest regression models were used to predict the percentage of root growth dry matter, while logistic regression and support vector machines were employed to classify crop production regions. To enhance model interpretability, the LIME tool was applied to analyze feature importance. The experimental results demonstrate that the models perform well in terms of prediction accuracy, and LIME provided clear feature explanations, helping identify the variables with the greatest impact on prediction outcomes. This research offers data-driven insights for optimizing resource management in smart agricultural systems.

Prioritization of Software Bugs Using Entropy‐Based Measures

ABSTRACTOpen‐source software is evolved through the active participation of users. In general, a user request for bug fixing, the addition of new features, and feature enhancements. Due to this, the software repositories are increasing day by day at an enormous rate. Additionally, user distinct requests add uncertainty and irregularity to the reported bug data. The performance of machine learning algorithms drastically gets influenced by the inappropriate handling of uncertainty and irregularity in the bug data. Researchers have used machine learning techniques for assigning priority to the bug without considering the uncertainty and irregularity in reported bug data. In order to capture the uncertainty and irregularity in the reported bug data, the summary entropy–based measure in combination with the severity and summary weight is considered in this study to predict the priority of bugs in the open‐source projects. Accordingly, the classifiers are build using these measures for different machine learning techniques, namely, k‐nearest neighbor (KNN), naïve Bayes (NB), J48, random forest (RF), condensed nearest neighbor (CNN), multinomial logistic regression (MLR), decision tree (DT), deep learning (DL), and neural network (NNet) for bug priority prediction This research aims to systematically analyze the summary entropy–based machine learning classifiers from three aspects: type of machine learning technique considered, estimation of various performance measures: Accuracy, Precision, Recall, and F‐measure and through existing model comparison. The experimental analysis is carried out using three open‐source projects, namely, Eclipse, Mozilla, and OpenOffice. Out of 145 cases (29 products X 5 priority levels), the J48, RF, DT, CNN, NNet, DL, MLR, and KNN techniques give the maximum F‐measure for 46, 35, 28, 11, 15, 4, 3, and 1 cases, respectively. The result shows that the proposed summary entropy–based approach using different machine learning techniques performs better than without entropy‐based approach and also entropy‐based approach improves the Accuracy and F‐measure as compared with the existing approaches. It can be concluded that the classifier build using summary entropy measure significantly improves the machine learning algorithms' performance with appropriate handling of uncertainty and irregularity. Moreover, the proposed summary entropy–based classifiers outperform the existing models available in the literature for predicting bug priority.

Exploring the Potential of Adaptive, Local Machine Learning in Comparison to the Prediction Performance of Global Models: A Case Study from Bayer's Caco-2 Permeability Database.

Machine learning (ML) techniques are being widely implemented to fill the gap in simple molecular design guidelines for newer therapeutic modalities in the extended and beyond rule of five chemical space (eRo5, bRo5). These ML techniques predict molecular properties directly from the structure, allowing for the prioritization of promising compounds. However, the performance of models varies greatly among ML use cases. A molecular property for which achieving sufficient performance in generalizing global models still remains difficult is Caco-2 permeability. Especially within the lower permeability ranges, which are specific for larger molecules belonging to the e/bRo5 space, accurate regression predictions have proven to be challenging. The present study, therefore, identifies a suitable combination of ML algorithm and descriptors, consisting of the LightGBM algorithm and RDKit molecular property descriptors, to predict Caco-2 permeability very efficiently by a simple global model. An additionally introduced local model uses the same algorithm and descriptors but selects its training data based on Tanimoto fingerprint similarity to match the individual test compound's structure. Evaluation of this adaptive model, by systematically varying the number of most similar structures for training, shows that, in comparison to the global model, there was only marginally improved performance with specific training data constellations. These random improvements indicate that deriving general rules for local model parametrization is not possible a priori for the chosen algorithm and descriptor combination, and preselecting training data does not seem advantageous over global ML based on all available data, while creation of more data-efficient models was generally proven to be possible.

Optimizing Agricultural Outcomes: Machine Learning in Crop Yield Prediction

Digital agriculture is becoming more and more in demand as a result of the quick advancement of information technology. Crop yield has always garnered a lot of interest as a significant problem in agricultural production. At the moment, machine learning and artificial intelligence in general are the most popular methods for predicting agricultural productivity. Consequently, one of the main problems in digital agriculture is creating a machine learning technique that can reliably forecast crop yield. Crop yield prediction has a strong time correlation, in contrast to conventional regression prediction problems. For instance, there are significant temporal correlations in the weather data for every county. Furthermore, crop yield is somewhat impacted by geographic data from various places. For instance, a county is likely to have large yields if its neighbouring countries have a strong harvest. We used models like random forest, decision tree classifier, support vector machine, KNN, and logic regression in this study. With an accuracy score of 99.77%, random forest produced the greatest results out of all of them.

Memristor Array Based on Wafer-Scale 2D HfS2 for Dual-Mode Physically Unclonable Functions.

Conventional Si-based physically unclonable functions (PUFs) take advantage of the unique variations in the fabrication processes. However, these PUFs are hindered by limited entropy sources and susceptibility to noise interference. Here we present a memristive device based on wafer-scale (2-in.) two-dimensional (2D) hafnium disulfide (HfS2) grown by molecular beam epitaxy (MBE). The polycrystalline HfS2 thin film can offer enhanced entropy sources for PUF applications, such as lattice defects, which can facilitate the random formation of conductive filaments and result in significant device-to-device (D2D) variations. Our proposed PUF design seamlessly integrates two distinct operating modes within a single circuit module. First, a reconfigurable and highly secure mode 1, and second, an ultrareliable mode 2, both with near-ideal Entropy (∼1.0), normalized Hamming distance (∼0.5) and correlation coefficient (∼0.0). Additionally, a predictive Fourier regression model further confirms the unpredictable nature of our dual-mode PUF, with an average prediction accuracy of ∼50%.

Research on prediction of surrounding rock deformation and optimization of construction parameters of high ground stress tunnel based on WOA-LSTM

To accurately understand the deformation behavior of surrounding rock in the jiugongshan No.2 high ground stress tunnel and optimize construction parameters for improved efficiency, on-site monitoring was used to gather data on rock deformation and initial support contact pressure. A WOA-LSTM regression prediction model was proposed, and excavation advance depth was optimized using numerical simulation. The WOA-LSTM regression prediction model demonstrated high accuracy in tunnel deformation monitoring. The absolute errors between predicted and actual values of tunnel settlement and convergence at various measurement points average 2.53− 04 mm and 1.96− 04 mm, with relative errors averaging 2.15% and 2.34%. These results meet the requirements for guiding construction. Additionally, based on the results of numerical simulation calculations, when the step length is 12 m during the construction of high-stress tunnels using the benching method, excavation advances of 2 m and 3 m result in settlement and convergence increases of 35.1% and 63.4%, and 25.5% and 55.2%, respectively, compared to an excavation advance of 1 m. However, no sudden jumps in these values were observed, indicating that with a 3 m excavation advance, the integrity of the surrounding rock remains in good condition, effectively guiding safe and rapid construction methods.

Diabetes Prediction Based on KNN, XGBoost, SVM and LR model

Abstract. A. Diabetes mellitus is a chronic metabolic disease characterized by high blood sugar levels due to insulin production problems or insulin resistance. Early identification of diabetes is crucial for preventing associated complications and effectively managing the condition. This study explores the application of four machine learning models, i.e., K-Nearest Neighbor (KNN), XGBoost, Support Vector Machine (SVM) and Logistic Regression (LR) in diabetes prediction. the main goal is to assess and contrast these models' efficacy in identifying diabetes risk, thereby helping healthcare professionals make timely diagnostic and treatment decisions. The results show that the logistic regression model with an AUC value of 0.95 performs much better than the other models, demonstrating excellent sensitivity and specificity in diabetes identification. The XGBoost model also demonstrates considerable predictive accuracy with an AUC value of 0.84, highlighting its ability to effectively handle large-scale datasets. Although the SVM and KNN models had slightly lower AUC values of 0.79, they still provided reliable predictive capabilities. These results demonstrate how machine learning may be used to improve diabetes prediction.

Rapid Raman spectroscopy analysis assisted with machine learning: a case study on Radix Bupleuri.

Radix Bupleuri has been widely used for its plentiful pharmacological effects. But it is hard to evaluate their safety and efficacy because the concentrations of components are tightly affected by the surrounding environment. Thus, Radix Bupleuri samples from different regions and varieties were collected. Based on the experimental and computational Raman spectrum, machine learning is emphasized for certain obscured characteristics; for example, linear discriminant analysis (LDA), support vector machine (SVM), eXtreme gradient boosting (XGBoost) and light gradient boosting machine (LightGBM). After dimension reduction by LDA, models of SVM, XGBoost and LightGBM were trained for classification and regression prediction of Bupleurum production regions. Support vector classifiers achieved the best accuracy of 98% and an F1 score above 0.96 on the test set. Support vector regression has a good fitting performance with an R2 score above 0.90 and a relatively low mean square error. However, complex models were prone to overfitting, resulting in poor generalization ability. Among these machine learning models, the typical LDA-SVM models, consistent with the high-performance liquid chromatography results, demonstrate great performance and stability. We envision that this rapid classification and regression technique can be extended to predictions for other herbs. © 2024 Society of Chemical Industry.

Regression prediction of tobacco chemical components during curing based on color quantification and machine learning

Color is one of the most important indicators to characteristic the quality of tobacco, which is strongly related to the variations of chemical components. In order to clarify the relationship between the changes of tobacco color and chemical components, here we established several prediction models of chemical components with the color values of tobacco based on machine learning algorithms. The results of correlation analysis showed that tobacco moisture content was highly significantly correlated with the parameters such as a*, H* and H°, the reducing sugar and total sugar content of tobacco was significantly correlated with the color values, and the starch content was highly significantly correlated with the color values except for b* and C*. The random forest models performed best in predicting tobacco moisture, reducing sugar, total sugar and starch constructed with the R2 of the model validation set was higher than 0.90, and the RPD value was greater than 2.0. The consistent between the predictions and measurements verified the availability and feasibility using color values to predict some chemical components of the tobacco leaves with high accuracy, and which has distinct advantages and potential application to realize the real-time monitoring of some chemical components in the tobacco curing process.

Soft sensor model for endpoint carbon content and temperature in BOF steelmaking based on just-in-time learning with multilayer structure preservation and sparse information enhancement

Ensuring precise prediction of the endpoint carbon content and temperature is paramount in controlling the endpoint of the basic oxygen furnace (BOF) steelmaking process. However, due to the frequent fluctuations in real-time blowing conditions, the conventional just-in-time learning soft sensor approach encounters challenges in effectively anticipating the blowing endpoint. To address these aforementioned issues, this research introduces a novel approach known as the pattern-oriented multilayer structure preservation and sparse information enhancement model (POMLSP-SIE). This approach involves dynamically generating a dataset with the same distribution based on the characteristics of the query sample pattern. It is combined with a multi-pathway dimension reduction model to preserve multi-view spatial geometry information while reducing data dimension. The low-dimensional embedding serves as dictionaries containing various hierarchical structural characteristics. Significant information within these dictionaries is sparsely represented to amplify its influence during the regression prediction process while diminishing the impact of less relevant information. This refinement aims to rectify the problem of static regression models being weak to adapt to changing working conditions, ultimately enhancing prediction performance. The effectiveness of the proposed method is substantiated through an experimental study utilising real converter steelmaking process data, thereby confirming its practical applicability.

Modeling of an inductive displacement sensor based on 1DCNN-LSTM-AT

Abstract The input-output relationship of inductive displacement sensors is typically nonlinear, which demands numerous computational resources for precise calculation. Therefore, the traditional analytical methods for the inductive displacement sensors cannot meet the real-time high-precision measurement requirements. In response to the aforementioned issues, this paper proposes an optimized Long Short-Term Memory (LSTM) neural networks based on one-dimensional convolutional neural networks (1DCNN) to modeling the input-output relationship of the inductive displacement sensor. First, the measurement principle of the inductive displacement sensors was analyzed, and the analytical model of the sensor output was derived. And the influences of the key parameters of sensors on the relationship between the induced voltage and the displacement were studied. Then, the 1DCNN-LSTM-AT network for modelling the input-output relationship of the inductive displacement sensor was studied. The spatial features of historical induced electromotive force (EMF) data generated by the induction coil were initially extracted using the 1DCNN network. And these spatial features were then utilized as input to the LSTM neural network to capture the temporal features of the historical induced EMF data. Subsequently, the spatiotemporal features of the induced EMF data were fed into the regression prediction layer to compute the displacement measurement results corresponding to the current input. Moreover, the attention mechanism was used for the 1DCNN-LSTM model to enhance the prediction accuracy and stability of the model. Finally, the experimental results demonstrate that the proposed 1DCNN-LSTM-AT model achieves an average absolute percentage error of 3.1%, significantly outperforming traditional models such as LSTM (29.3%), CNN (4.7%), and ANN (4.3%). This paper provides a new method for modelling the nonlinear relationships of the inductive displacement sensor, and presents a fresh perspective for research in the data processing of nonlinear sensors.

Forest Aboveground Biomass Estimation Based on Unmanned Aerial Vehicle-Light Detection and Ranging and Machine Learning.

Eucalyptus is a widely planted species in plantation forests because of its outstanding characteristics, such as fast growth rate and high adaptability. Accurate and rapid prediction of Eucalyptus biomass is important for plantation forest management and the prediction of carbon stock in terrestrial ecosystems. In this study, the performance of predictive biomass regression equations and machine learning algorithms, including multivariate linear stepwise regression (MLSR), support vector machine regression (SVR), and k-nearest neighbor (KNN) for constructing a predictive forest AGB model was analyzed and compared at individual tree and stand scales based on forest parameters extracted by Unmanned Aerial Vehicle-Light Detection and Ranging (UAV LiDAR) and variables screened by variable projection importance analysis to select the best prediction method. The results of the study concluded that the prediction model accuracy of the natural transformed regression equations (R2 = 0.873, RMSE = 0.312 t/ha, RRMSE = 0.0091) outperformed that of the machine learning algorithms at the individual tree scale. Among the machine learning models, the SVR prediction model accuracy was the best (R2 = 0.868, RMSE = 7.932 t/ha, RRMSE = 0.231). In this study, UAV-LiDAR-based data had great potential in predicting the AGB of Eucalyptus trees, and the tree height parameter had the strongest correlation with AGB. In summary, the combination of UAV LiDAR data and machine learning algorithms to construct a predictive forest AGB model has high accuracy and provides a solution for carbon stock assessment and forest ecosystem assessment.

Landslide Displacement Prediction Stacking Deep Learning Algorithms: A Case Study of Shengjibao Landslide in the Three Gorges Reservoir Area of China

Computational models enable accurate, timely prediction of landslides based on the monitoring data on-site as the development of artificial intelligence technology. The most existing prediction methods focus on finding a single prediction algorithm with excellent performance or an integrated and efficient hyperparameter optimization algorithm with a highly accurate regression prediction algorithm. In order to break through the limitation of generalization of prediction models, this paper proposes an ensemble model that combines deep learning algorithms, with a stacking framework optimized with the sliding window method. Multiple deep learning algorithms are set as the first layer of the stacking framework, which is optimized with the sliding window method to avoid confusion in the time order of datasets based on time series analysis. The Shengjibao landslide in the Three Gorges Reservoir is used as a case study. First, the cumulative displacement is decomposed into a trend and a periodic term using a moving average method. A single-factor and a multi-factor superposition model based on multiple deep learning algorithms are used to predict the trend and periodic term of the displacement, respectively. Finally, the predicted values of the trend and periodic terms are added to obtain the total predicted landslide displacement. For monitoring point ZK2-3, the values of RMSE and MAPE of the total displacement prediction with the stacking model are 15.93 mm and 0.54%, and the values of RMSE and MAPE of the best-performing individual deep learning model are 20.00 mm and 0.64%. The results show that the stacking model outperforms other models by combining the advantages of each individual deep learning algorithm. This study provides a framework for integrating landslide displacement prediction models. It can serve as a reference for the geological disaster prediction and the establishment of an early warning system in the Three Gorges Reservoir Area.

A Temporal–Geospatial Deep Learning Framework for Crop Yield Prediction

With the rapid development of information technology, the demand for digital agriculture is increasing. As an important agricultural production topic, crop yield has always attracted much attention. Currently, artificial intelligence, particularly machine learning, has become the leading approach for crop yield prediction. As a result, developing a machine learning method that accurately predicts crop yield has become one of the central challenges in digital agriculture. Unlike traditional regression prediction problems, crop yield prediction has a significant time correlation. For example, weather data for each county show strong temporal correlations. Moreover, geographic information from different regions also impacts crop yield to a certain extent. For example, if a county’s neighboring counties have a good harvest, then this county is likely to have high yields as well. This paper introduces a novel hybrid deep learning framework that combines convolutional neural network (CNN), graph attention network (GAT) and long short-term memory (LSTM) modules to enhance prediction accuracy. Specifically, CNN is employed to extract the features from the input data for each county in each year. GAT is introduced to model the geographical relationships between neighboring counties, allowing the model to capture spatial dependencies more effectively. LSTM is used to extract the temporal information within many years. The proposed hybrid deep learning framework CNN-GAT-LSTM captures both the temporal and spatial relationships, thereby improving the accuracy of yield prediction. We conduct experiments on a nationwide dataset that includes data from 1115 soybean-producing counties in 13 states in the United States covering the years from 1980 to 2018. We evaluate the performance of our proposed CNN-GAT-LSTM model based on three metrics, namely root of the mean squared error (RMSE), R-squared (R2) and correlation coefficient (Corr). The experimental results demonstrate that the proposed model achieves significant performance improvements over the existing state-of-the-art model, with RMSE reduced by 5%, R2 improved by 6% and Corr enhanced by 4%.

Left ventricular late diastolic strain rate predicts cardiovascular mortality in patients with heart failure with reduced ejection fraction

Abstract Background Left ventricular (LV) strain rate measures provide detailed information regarding LV contraction and filling pressures. These include global systolic strain rate (GSRs), early and late diastolic strain rates (GSRe and GSRa), and ratio of transmitral early filling velocity to GSRe (E/GSRe). Purpose To investigate the prognostic value of strain rate measures with regard to cardiovascular mortality in patients with heart failure with reduced ejection fraction (HFrEF). Methods The study population was retrospectively included from a HF clinic and consisted of patients with LV ejection fraction (LVEF) ≤45%. Echocardiography was obtained within 1 year of referral. The study endpoint was cardiovascular mortality, which was analysed using univariable and adjusted Cox proportional hazards models. Incremental prognostic information was assessed using Harrell’s C-index and continuous net reclassification improvement (NRI). Cubic spline analysis was used to dichotomise variables at a hazard ratio (HR) of 1 for better interpretability. Results The final study population consisted of 1160 patients with HFrEF (mean age 68±11.8 years, 72.7% male, LV ejection fraction (LVEF) 28.6±9.9%). During follow-up (median 10.1 years, IQR: 4.7-13.0), 333 (28.7%) met the outcome. Subjects who met the outcome generally had higher age and prevalence of male sex, a higher burden of comorbidities, and more severely impaired conventional echocardiographic parameters. All four strain rate measures were significantly associated with cardiovascular mortality (GSRs: HR 16.88 95% CI [8.42, 33.84] per 1 s-1 increase, GSRe: HR 0.29 [0.18, 0.47] per 1 s-1 increase, E/GSRe ratio: HR 1.31 [1.22, 1.41] per 1 unit increase, GSRa: HR 0.15 95% CI [0.23, 0.65] per 1 s-1 increase, p<0.0001 for all). However, only GSRa remained an independent predictor in multivariable Cox regression after adjustment for clinical risk factors including sex, age, mean arterial pressure, heart rate, body mass index, cigarette pack years, diabetes mellitus, total cholesterol, ischemic heart disease, and echocardiographic factors including LVEF, global longitudinal strain, tricuspid annular plane systolic excursion, LV mass index, LV internal diameter index, E/e’ ratio and left atrial volume index (HR 0.39 [0.23,0.65], p=0.0003). Furthermore, GSRa provided incremental prognostic information to a model containing all aforementioned parameters (Harrell’s C-index 0.789 vs. 0.792 and continuous NRI 0.158). GSRa was dichotomised at 0.5 s-1, and a Kaplan-Meier curve stratified according to this was plotted (log rank p<0.0001)(Figure 2). Conclusion GSRa is an independent predictor of cardiovascular mortality in patients with HFrEF after multivariable adjustment. GSRa provides incremental prognostic value over conventional echocardiographic measures in risk stratification of HFrEF patients. Our results suggest an increased risk of cardiovascular mortality associated with a GSRa below 0.5 s-1.Figure 1:Spline of HR according to GSRaFigure 2:Kaplan-Meier curve

The impact of diabetes on cardiac remodelling after transcatheter aortic valve implantation

Abstract Background Transcatheter aortic valve implantation (TAVI) induces reverse cardiac remodelling, an important outcome associated with lower mortality. Diabetes mellitus (DM) is common among TAVI patients and is one of the fastest growing chronic conditions worldwide. Despite its propensity to adversely affect myocardium, the influence of DM on post-TAVI remodelling remains poorly understood. Purpose To characterise the influence of DM on cardiac remodelling following TAVI. Methods Retrospective analysis of 202 patients undergoing TAVI at an Australian tertiary centre. 12-month myocardial remodelling was compared between DM and no DM groups. Results Of 202 patients, median age 82 (IQR: 76, 86) years, 50% female and median STS score 3.5 (2.5, 4.9), 62 (31%) had Type II DM. A robust multivariable analysis (C-Statistic 0.83) revealed DM (adjusted odds ratio [aOR] 2.3, p=0.042) and ≥ mild post-procedural paravalvular regurgitation (aOR 2.5, p=0.027) were independent predictors of poor left-ventricular mass index regression (LVMIr). Among diabetic patients, suboptimal disease control was associated with inferior LVMIr (6% vs -9%, p=0.011), while insulin dependency was not (-3% vs 2%, p=0.489). At 12-months, no differences were found in left-atrial remodelling, left and right-ventricular function, and prosthetic valve performance between DM and no DM groups. Poor LVMIr was associated with lower survival in Kaplan Meier analysis (65% vs 85%, log-rank p=0.049). Conclusion Among patients undergoing TAVI, those with DM demonstrate poorer LVMIr at 12-months. Improved glycaemic control may serve as an important strategy to optimise myocardial remodelling in diabetic patients undergoing TAVI.12-Month LVMI Regression

Machine learning to predict incident atrial fibrillation in patients with acute stroke: FIND-AF STROKE

Abstract Background Atrial fibrillation (AF)-related strokes have a high rate of recurrence, and are associated with substantial morbidity, disability, healthcare expenditure, and mortality.(1-3) Detection of AF promises opportunity to prevent recurrent stroke through conversion of antiplatelet to anticoagulant therapy,(4,5) but use of extended monitoring has cost implications.(6) Thus accurately identifying patients with stroke at high risk for AF is important but previous previous models have been developed in small, unrepresentative cohorts.(7) Purpose To develop and validate a simple novel decision support tool for prediction of incident AF after hospitalisation with stroke through machine learning. Methods We developed the model (FIND-AF STROKE) in UK secondary care and death certificate-linked primary care health record data from individuals aged ≥30 years without known AF hospitalized with a stroke or transient ischaemic attack (Jan 2, 1998 to Feb 28, 2022), randomly divided into training (80%) and testing (20%) datasets. We evaluated logistic regression and random forest models for prediction of 2-year AF risk in the testing dataset with internal bootstrap validation, and compared against the CHA2DS2VASC score. Youden index was used to set a threshold for higher and lower risk. Results Of 35 530 UK individuals (mean age 72.1 (SD 12.7) years, 50.9% women), 1742 (4.8%) had developed incident AF at 2 years. Using feature importance a parsimonious version of the model was created requiring only age, sex, ethnicity (white versus other) and five comorbidities. In the testing dataset prediction performance for the random forest algorithm (area under the receiver operating characteristic (AUROC) 0.710, 95% CI 0.685 – 0.736) was superior to the CHA2DS2VASC score (0.645, 95% CI 0.617-0.672) (Figure 1). The threshold for higher risk using the Youden index was 4.81% which resulted in a negative predictive value of 97.6%, sensitivity 0.712, specificity 0.583. Of the post-stroke patients 44% were labelled at higher risk of AF, who were older, had a higher prevalence of comorbidities, and a 3.3-fold higher incidence of AF post-stroke, than those identified as lower risk (Table 1). Conclusions The machine learning FIND-AF STROKE decision support tool can accurately identify individuals at risk of incident AF after stroke to enable a targeted and cost-effective approach to extended monitoring. Validation within cohorts that have undergone extended monitoring post-stroke will give a more accurate estimation of yield at different risk thresholds.Figure 1