Performance Of Model Research Articles

Machine learning model using cardiovascular magnetic resonance to predict cardiovascular events in asymptomatic patients with known CAD

Abstract Background Despite declining rates, recurrent cardiovascular events continue to pose significant mortality and morbidity risks in patients with obstructive coronary artery disease (CAD). Traditional prognostic assessments in these patients rely on limited clinical and imaging data. Although several studies have shown the strong prognostic value of stress cardiovascular magnetic resonance (CMR) in asymptomatic patients with known CAD, no dedicated score using CMR was established. Machine learning (ML) methods present an opportunity to leverage a wider array of variables to address this gap. Purpose This study sought to investigate the feasibility and accuracy of ML using stress CMR and clinical data to predict major adverse cardiovascular events (MACE) over a 10-year period in patients with known obstructive CAD, comparing its performance with traditional statistical methods. Methods Between 2009 and 2011, consecutive asymptomatic patients with known obstructive CAD referred for vasodilator stress CMR in two independent centres were followed for the occurrence of MACE defined by cardiovascular mortality or recurrent non-fatal myocardial infarction (MI). Known obstructive CAD was defined by a history of percutaneous coronary intervention, coronary artery bypass surgery, or MI. The first center (N=866) was designated as the derivation cohort for variable selection and model building. The second center (N=224) was used as the validation cohort to evaluate the performance of the final ML-model. Various feature selection methods (Random Forest [RF], XGBoost, and Boruta) were employed to identify robust and informative variables. A final RF model was trained on the derivation cohort and evaluated on the validation cohort. Performance was compared with a standard generalized logistic regression model (GLM), utilizing the receiver operating characteristics (ROC) and the precision-recall (PR) curves as well as the area under the curves (AUC). Results In the derivation cohort (age 67±7 years, LVEF 43±8%, 81% male), 243 (28%) out of 866 patients encountered MACE during the 10-year follow-up period while in the validation cohort (age 67±7 years, LVEF 43±7%, 84% male), 53 (24%) out of 224 patients experienced MACE. Through feature selection method in the derivation cohort, one clinical and four CMR variables emerged as key predictors: number of ischemic segments, number of late gadolinium enhancement (LGE) segments, left ventricular ejection fraction, left ventricular end-diastolic diameter indexed, and age (Figure 1). Using those 5 variables, the RF model significantly outperformed the GLM, demonstrating superior predictive accuracy (AUROC: 0.89 vs. 0.73; PRAUC: 0.78 vs. 0.58, all p<0.001) in the validation cohort (Figure 2). Conclusions A supervised machine learning model using RF, integrating age and stress-CMR data, outperformed traditional statistical methods in predicting 10-year MACE for asymptomatic patients with known CAD.Figure1:Variable selectionFigure2:Models performance on validatio

Can a zero-shot learning Large Language Model code complex interview data?

Abstract Background Psychological autopsy is essential to establish theories, explore trends and identify previously unexplored psychosocial risk factors in suicide research. However, qualitative research has been scrutinized for being prone to interpretation bias, problems with accuracy, challenges to reproducibility, and is very time- and cost intensive. The current study aimed to investigate if a Large Language Model can achieve sufficient agreement with a researcher in the deductive coding of interview data obtained in a psychological autopsy study of suicide, to be integrated with qualitative research procedures. Methods Data from 38 interviews was deductively coded by a researcher and a LLAMA 3 based language model. The model performance was evaluated in four increasingly difficult coding tasks, including binary classification and data summarization. Intercoder agreement scores were calculated using Cohen’s Kappa. Results The preliminary results showed that the LLM achieved substantial agreement with the human coders for the binary classification task (.78). The variability in performance was influenced by code definitions. The results of the quality of LLM interpretation and summarization will also be presented. Conclusions State-of-the-art LLM can be easily integrated into the qualitative analyses of psychological autopsy interviews and may improve real time monitoring of suicides. We recommend a human-AI collaborative model, whereby deductive coding by the LLM is complemented by human inductive coding and further interpretation. Key messages • Integrating an LLM with qualitative research procedures is feasible in a collaborative model. • Integrating an LLM with qualitative research procedures allows near real-time monitoring based on qualitative data, which extends to other public health fields.

Adding ethnicity to cardiovascular risk prediction: external validation and model updating of SCORE2 using data from the multiethnic HELIUS population cohort

Abstract Background Current prediction models for mainland Europe do not include ethnicity, despite ethnic disparities in cardiovascular disease (CVD) risk. Purpose We evaluated the performance of SCORE2 across the largest ethnic groups in the Netherlands with and without the factor of ethnicity. Methods We linked data from 11,614 participants enrolled between 2011 and 2015 in the prospective, multi-ethnic HELIUS cohort study, the Netherlands, aged between 40 and 70 years without CVD with hospital data. Variables (age, sex, smoking status, blood pressure, cholesterol) and outcome first CVD event (myocardial infarction, stroke, CVD death) were consistent with SCORE2. We used Fine and Gray models to calculate sub-distribution hazard ratios (SHR) for South-Asian Surinamese, African Surinamese, Ghanaian, Turkish and Moroccan origin groups, relative to the Dutch origin group, on top of individual SCORE2 risk predictions. Model performance was evaluated by discrimination, calibration and net reclassification index (NRI). Results Overall, 274 fatal and non-fatal CVD events, and 146 non-cardiovascular deaths were observed during a median of 7.8 years follow-up (IQR 6.8–8.8). SHRs for CVD events were 1.86 (95% CI 1.31-2.65, p=0.0005) for the South-Asian Surinamese, 1.09 (95% CI 0.76-1.56) for the African-Surinamese, 1.48 (95% CI 0.94-2.31) for the Ghanaian, 1.63 (95% CI 1.09-2.44, p=0.018) for the Turkish, and 0.67 (95% CI 0.39-1.18) for the Moroccan origin groups. Adding ethnicity to SCORE2 led to comparable calibration and discrimination [0.76 (0.74-0.79) vs. 0.77 (95% CI 0.74-0.78)]. The NRI for adding ethnicity to SCORE2 was 0.24 (95% CI 0.18-0.31) for events and -0.12 (95% CI -0.13--0.12) for non-events. Conclusion Adding ethnic backgrounds improves risk classification of the SCORE2 risk prediction model in a middle-aged, multi-ethnic Dutch population, which may help to address disparities in CVDs through timely treatment.

Understanding and monitoring vulnerability during pregnancy at population-level

Abstract Background Detecting and addressing vulnerability during pregnancy is important to optimize the first thousands days of children’s lives and prevent health inequities. We aimed to gain insight into vulnerability during pregnancy and determine the best data to predict vulnerability. Methods In the Netherlands, two studies were done using nationwide routinely collected data and self-reported Public Health Monitor-data. The first study utilized all data to identify classes of vulnerability among 4172 pregnant women through Latent Class Analysis. The second study predicted multidimensional vulnerability at population-level using solely routinely collected data first, and step-by-step incorporating self-reported data. Random Forest (RF) analysis was employed, and model performance was assessed using Area Under the Curve (AUC) and F1-measure. Results Five vulnerability-classes were found: ‘healthy and socioeconomically stable’, ‘high care utilization’, ‘socioeconomic vulnerability’, ‘psychosocial vulnerability’, and ‘multidimensional vulnerability’. Women in the ‘multidimensional vulnerability’-class shared multiple risk factors across domains and lacked protective factors. They more often had adverse outcomes such as premature birth. In the next study, the initial RF-model achieved an AUC of 0.98 in distinguishing multidimensional vulnerability, with an F1-measure of 0.70. Adding self-reported data improved model performance. Strong predictors were self-reported health, socioeconomic characteristics and healthcare utilization. Conclusions Several vulnerability classes can be distinguished among pregnant women. The co-existence of risk factors and lack of protective factors leads to more adverse outcomes. Effective public health strategies should include both medical and social care and support. Next, it seems possible to predict multidimensional vulnerability using data that is readily available, providing a foundation for longitudinal monitoring and policy formulation. Key messages • Considering the combination of both social risk and protective factors related to vulnerability during pregnancy is necessary. • Routinely collected data could provide insight in the prevalence, geographical distribution and trends in multidimensional vulnerability during pregnancy at population-level.

Predicting major adverse cardiovascular events in suspected myocarditis patients using cmr late gadolinium enhancement radiomics: a time-to-event prognostication modeling

Abstract Background Myocarditis, characterized by inflammation of the heart tissue, exhibits a wide range of clinical manifestations; however, stratifying its risk is difficult. Objective We aimed to prognosticate Major Adverse Cardiovascular Events (MACE, i.e., death, sustained ventricular arrhythmia, heart failure hospitalization, and recurrent myocarditis) in patients with suspected myocarditis using cardiac magnetic resonate (CMR) late gadolinium enhancement (LGE) radiomic features. Method We enrolled consecutive 1099 patients (age: 47.68±16.56, 39.63% female) undergoing CMR with the initial suspicion of myocarditis based on the judgment of the referring physician, recruited from two major registries. We used fully automated deep learning tools to determine the segmental extent of LGE in the left ventricle. Furthermore, we pre-processed the dataset through bias-field correction and intensity normalization to ensure the model consistency across various image acquisitions. Different radiomic features, including shape, intensity, and texture, were extracted from CMR-LGE. We employed several feature selection methods and time-to-event survival models based on radiomics features. The development process, covering pre-processing, feature selection, and hyperparameter optimization, utilized 70% of the dataset (10-fold cross-validation). The evaluation was performed on the remaining 30% of the test set. Model performance was measured using the c-index and cumulative AUC (C-AUC) on the untouched 30% test set. Results The top-performing model, employing a 90% variance threshold with the survival support gradient boosting model, achieved a C-Index of 0.76 and a cumulative C-AUC of 0.80. The Random Forest Survival model was a close second, with a C-Index of 0.75 and a C-AUC of 0.79. The conventional multivariate Cox-PH model achieved a C-Index of 0.67 and a C-AUC of 0.70. Conclusion Utilizing radiomics features from the CMR-LGE, combined with machine learning algorithms, allows accurate prediction of MACE in suspected myocarditis patients. The performance of the Radiomics model could be improved by including other modality information.

Using deep learning to predict cardiovascular magnetic resonance findings from echocardiography videos

Abstract Background Echocardiography is the most common modality for assessing cardiac structure and function. While cardiac magnetic resonance (CMR) imaging is less accessible, CMR can provide unique tissue characterization including late gadolinium enhancement (LGE), T1 and T2 mapping, and extracellular volume (ECV) which are associated with tissue fibrosis, infiltration, and inflammation. While deep learning has been shown to uncover findings not recognized by clinicians, it is unknown whether CMR-based tissue characteristics can be derived from echocardiography videos using deep learning. Purpose To assess the performance of a deep learning model applied to echocardiography to detect CMR-based parameters. Methods In a retrospective single-center study, adult patients with CMRs and echocardiography studies within 30 days were included. A video-based convolutional neural network was trained on echocardiography videos to predict CMR-derived labels including WMA presence, LGE presence, and abnormal T1, T2 or ECV across echocardiography views. The model performance was evaluated in a held-out test dataset not used for training. Results The study population included 1,453 adult patients (mean age 56±18 years, 42% female) with 2,556 paired echocardiography studies occurring on average 2 days after CMR (interquartile range 2 days prior to 6 days after). The model had high predictive capability for presence of WMA (AUC 0.873 [95%CI 0.816-0.922]), however, the model was unable to reliably detect the presence of LGE (AUC 0.699 [0.613-0.780]), native T1 (AUC 0.614 [0.500-0.715]), T2 0.553 [0.420-0.692], or ECV 0.564 [0.455-0.691]). Conclusions Deep learning applied to echocardiography accurately identified CMR-based WMA, but was unable to predict tissue characteristics, suggesting that signal for these tissue characteristics may not be present within ultrasound videos.Central Figure -Study Pipeline-

Digitisation of electrocardiographic images enable artificial intelligence-based mortality prediction

Abstract Background Artificial intelligence-based electrocardiogram (AI-ECG) models have shown excellent performance in a vast number of diagnostic and prognostic tasks. One of such tasks, which is critical for patient risk stratification and clinical intervention, is the prediction of adverse events, including death. Although a plethora of existing AI-ECG models have shown their accuracy in mortality prediction, their development and utilization currently rely on digital ECG signals for training and application. Purpose Many hospitals around the world have a significant number of ECGs stored in image formats or printed as paper ECGs. To address this problem, we developed and tested an ECG image digitisation pipeline to extract digital traces from ECG images and compare the performance of AI-ECG models for mortality prediction, under both natively digital and digitised signal formats. Methods A secondary care dataset of over 1 million ECGs (1,163,401 ECGs; 189,539 patients) was used for this analysis, in which both a natively digital signal and a PDF image was available. The ECG digitisation pipeline comprised of coloured PDF image inputs (2200 x 1700), initially processed to extract the ECG traces and grid features. ECG signals were extracted from the traces after cropping/restoration of the images and scaled appropriately based on the detected grid (2.5-sec strips per lead, 0.5 mV per grid block). Subsequently, both digital and digitised signals were identically pre-processed in order to acquire the ECG data fed into the AI model. In particular, a previously successful convolutional neural network architecture with a discrete-time survival loss function was used for mortality prediction. The model was trained under multiple configurations, using either the digital or the digitised ECGs as training/testing datasets (Figure 1). ECGs were split by patient ID to ensure cross-subject generalization with a training/validation/testing split of 50%/10%/40% ratio, respectively. Finally, the models were evaluated by the c-index. Results A comparison of the digital and digitised ECGs on the respective 2.5-sec strips available in both formats revealed a mean absolute error of 0.014 (CI: 0.07 – 0.021) and a Pearson’s R of 0.98 (CI: 0.96 – 1.00) between the signals, demonstrating an accurate ECG digitisation (Figure 2 shows overlapping natively digital (black) and digitised (red) traces). The performance of the digitisation was also evident in the performance of the AI-ECG mortality prediction task, with the model achieving c-indices of 0.775 (CI: 0.774 – 0.776) and 0.772 (CI: 0.771 – 0.774) when trained and tested on digital and digitised signals, respectively. Conclusion We described the first digitisation-based AI-ECG model for mortality prediction. Our results demonstrate the potential for clinical settings that lack the digital ECG infrastructure, to develop and deploy AI models using ECG image formats.Figure 1Figure 2

Comprehensive machine learning models for prediction of heart failure in 476,393 women and men from the UK Biobank reveal sex differences and underutilized risk factors

Abstract Background Machine learning (ML) models provide potential advantage over ‘traditional’ regression models in heart failure (HF) prediction. Objective To compare performances of Cox PH models and ML survival models for incident HF in men and women without prevalent ischemic heart disease (IHD). We also aimed to identify potential high-risk precursors otherwise ignored by conventional survival models, and to investigate differences between sex-specific models. Methods We included 476,393 participants (55.6% women) from the UK Biobank, after excluding participants with a history of HF or IHD, and defined sex-specific datasets. We predicted incident HF events using over 400 baseline characteristics. We constructed multivariable Cox PH models, which included all predictor variables and subsequently only those remaining after LASSO stability selection. We also developed two supervised ML models (Random Survival Forest (RSF), eXtreme Gradient Survival Boosting (XGBoost)). We identified the 15 most important sex-specific predictors in each model and performances were compared using the C-index. Models were validated using hold-out sets. Results During 12.3 ± 1.9 years of follow-up, 4680 (1.76%) women and 6631 (3.14%) men developed incident HF. XGBoost showed the best performance during model training (C-index, training: 0.89 in men, 0.97 in women; validation 0.77 in men, 0.80 in women). The multivariable Cox model performed second-best (C-index, training: 0.78 in men, 0.82 in women; validation: 0.76 in men, 0.78 in women). RSF performed slightly worse (C-index, training: 0.75 in men, 0.79 in women; validation: 0.75 in men, 0.79 in women) but did not show performance drop during validation. LASSO stability selection performed similar to RSF. Age, self-reported lifetime treatments and medications, cystatin-C, waist circumference and FEV1-scores were identified as strong risk factors in all models for both sexes. Reduced albumin levels and elevated HbA1c were more strongly associated with high risk in men, while elevated systolic BP showed higher importance in women. Traditional Cox models observed CRP as important only in men, while the ML models identified CRP as important for both sexes. Neutrophil count was considered a strong risk factor in both sexes in the traditional Cox models, yet it was not among the most important predictors in both ML models. Presence of other heart disease (which included a.o. pericardial disease, valve disorders and arrhythmias) was an important predictor variable only in the ML models. Conclusion ML models showed similar performance to Cox PH models for HF prediction. Despite this, differences in predictor importance were identified between models. Sex-specific risk predictors were found, and FEV1 score, which is not commonly included in existing models, was identified as an important risk factor. These results suggest that ML models may reveal additional insights that would otherwise remain unnoticed.

Clinical prediction model of main adverse cardiovascular and cerebrovascular events for elderly patients with acute coronary syndrome

Abstract Background Age is an independent factor for adverse outcomes in acute coronary syndrome (ACS). With the accelerated aging process, there is an increasing incidence of ACS in the elderly; however, research on this population is underrepresented. Due to the unique presence of multiple diseases and organ changes in the elderly, specialized risk assessment tools are required. Purpose To explore the risk factors for poor prognosis of elderly patients with ACS and establishing a risk stratification tool. Methods The derivation cohort consisted of 1674 elderly ACS patients consecutively enrolled in a single center from January 2013 to December 2013. The validation cohort included 2333 elderly ACS patients consecutively enrolled in nine medical centers from January 2015 to May 2019. All patients were followed up for 2 years, and the study endpoint was defined as major adverse cardiovascular and cerebrovascular events (MACCE). Model selection was performed using Cox regression and the Akaike Information Criterion (AIC) to construct the final risk assessment model. The performance of the predictive model was evaluated in the derivation and validation cohorts using the c-index, receiver operating characteristic (ROC) curves, and calibration curves. Results A total of 1,674 elderly ACS patients (70.97±4.68 years, 23.24% female) were included in the derivation cohort, and 210 cases (12.54%) of MACCE were recorded during the 2-year follow-up period. In the validation cohort, there were 2,333 elderly ACS patients (72.13±5.82 years, 35.83% female), with 364 cases (15.60%) of MACCE documented during the 2-year follow-up. Initially, 25 variables were preliminarily selected based on the results of univariate Cox regression and the clinical importance of these variables. Subsequently, these variables were included in a multivariable Cox model, and model selection was performed based on the AIC, resulting in the final selection of 10 variables (including age, cardiac insufficiency, current smoker, DAPT use, history of PCI or CABG, IABP use, number of lesions, residual SYNTAX score, preoperative HGB, and preoperative PLT). Then we developed and validated a nomogram to predict the 2-year MACCE in elderly ACS patients based on the assessment of 10 clinically easily obtainable variables during hospitalization. In both the derivation and validation cohorts, the nomogram demonstrated good discriminative ability (AUC of 0.723 and 0.699, c-index of 0.727 and 0.661, respectively) and calibration (calibration curves closely approximating the 45-degree diagonal line). The Kaplan-Meier analysis reveals significant differences in MACCE incidence rates between groups stratified by the median predicted probability (p<0.001). Conclusion Our study developed and validated a practical nomogram for predicting the 2-year risk of MACCE in elderly patients with ACS.

Sex-specific prediction of cardiogenic shock complicating acute coronary syndromes: the SEX-SHOCK score

Abstract Background Mortality of cardiogenic shock (CS) complicating acute coronary syndromes (ACS) remained nearly unchanged over the last two decades. Female CS patients are less likely to receive guideline recommended care, while having worse in-hospital outcomes as compared to males.1 Presently available tools to assess the risk of developing CS in the setting of ACS, such as the ORBI risk score, were developed in predominantly male patient populations, while not considering inflammatory mediators or proxies of cardiorenal function. Purpose The present study aimed (i) to test the sex-specific performance of the ORBI model to predict in-hospital CS in patients presenting with ACS, and (ii) to develop and externally validate a better performing risk prediction model for both, women and men. Methods Among 44’220 ACS patients recruited in the Swiss AMIS-Plus study, the sex-specific performance of the ORBI risk score to estimate the probability for the development of CS was tested. By harnessing regression- and machine-learning based modelling approaches (ie, random forest [RF], multiple layer perceptron [MLR], and logistic regression [LR]), independent predictors of in-hospital CS were identified in sex-disaggregated data. Best performing models and variables were then used to develop SEX-SHOCK, a novel risk prediction model that accounts for sex-specific disease characteristics. External validation was done in 4’787 ACS patients recruited in the SPUM-ACS study. Results The ORBI risk prediction model demonstrated lower discriminatory performance in female patients relative to males (AUC [95%CI]: 0.76 (0.74-0.78) vs. 0.81(0.80-0.83); p<0.001). RF, MLR, and LR identified C-reactive protein, creatinine, left-ventricular function, and ST-segment elevation as most potent predictors of in-hospital CS beyond the parameters included in ORBI, with the performance of LR-based models superseding those of RF and MLR. By combining most important predictors with best performing models, the SEX-SHOCK score was developed, showing superior performance as compared to ORBI in both females (AUC 0.80, 95%CI 0.78-0.82; p<0.01) and males ACS patients (AUC 0.83, 0.82-0.85; p<0.001), which prevailed upon 10-fold cross validation. Following external validation in SPUM-ACS, SEX-SHOCK showed improved discriminatory performance over ORBI in both females (0.84, 95% CI 0.78-0.89; p<0.05) and males (0.81, 95% CI 0.78-0.85; p<0.01), with the most pronounced improvement observed in female patients (△C-statistics%=6.23 vs. 2.79%; p <0.001). Conclusion Our study highlights important sex differences in the performance of available tools for risk prediction of developing in-hospital CS among patients presenting with ACS. By harnessing regression- and machine-learning-based approaches, the SEX-SHOCK risk score was developed, outperforming existing risk prediction models in internal and external validation cohorts.

Sex-specific models to predict 5-year mortality after ST-elevation myocardial infarction using machine learning: insight from FAST-MI registry

Abstract Background Most prognostic stratification tools in acute myocardial infarction (AMI) have been derived from populations including both women and men and including only a small proportion of women. Purpose Using supervised machine learning (ML), we examined the performance of 2 models assessing 5-year mortality after AMI, one derived from a population of women only and one from a population of men only, to determine whether sex-specific models improved outcome prediction. Methods This cohort study used the French registry on acute ST-elevation and non-ST-elevation myocardial infarction (FAST-MI) 2010 and 2015 surveys. This multicentric registry led in more than 200 French hospitals, enrolled all consecutive patients with acute myocardial infarction during a 1-month recruitment period. Our analysis included all men and women presenting STEMI who underwent invasive coronary angiography (ICA). To build and validate the models, the data set were split with a 70%/30% ratio into training and testing sets. The primary outcome was 5-year all-cause mortality. Then, 52 clinical, laboratory, ECG, echocardiographic, and ICA parameters were evaluated for feature selection using Boruta algorithm. Different supervised machine learning algorithms, including random forest (RF), were assessed for model building, and their performance were compared in women and men. To compare prediction accuracy according to sex, we evaluate the performance of the women-based ML model in men from the same cohort. Results 1,189 consecutive women and 3,685 men with STEMI (mean age 61±13 and 69±15 years, respectively) were recruited; 12% of men and 20% of women experienced 5-year all-cause mortality. Using Boruta algorithm, the 10 most important variables for prediction were selected (Figure 1). For women-based ML model building, the RF algorithm exhibited the best performance to predict mortality with an area under the receiver-operating characteristic curve (ROC-AUC) of 0.82 (95% CI: 0.77 – 0.88), an area under the precision-recall curve (PR-AUC) of 0.59 (95% CI: 0.54 – 0.64); and a F1-score of 0.58. The women-based ML model exhibited lower performance in men (ROC-AUC: 0.78; PR-AUC: 0.43; Figure 2). Conversely, the men-based model exhibited better accuracy in men than in women. Conclusion In a large multicentric cohort of STEMI patients, a women-based ML-model exhibited a good accuracy to predict 5-year all-cause mortality in women, but a drop of accuracy when applied to men. In mirror, the accuracy of the men-based model was lower in women. This suggests that sex-specific models might be superior to general models to predict mortality after AMI.Features selection and ML model buildingML model performance according to sex

Improved prognosis prediction of idiopathic pulmonary arterial hypertension using targeted plasma proteomics in disease management

Abstract Background Idiopathic pulmonary arterial hypertension (IPAH), a special subgroup of PAH with unclear etiology, is characterized by rapid progression and poor prognosis underscoring the need for optimization of IPAH management and the accurate prediction of prognosis. However, it is disappointing that specific plasma biomarkers for IPAH management and prognosis are lacking in clinical practice. Up to date, the plasma biomarkers used for assessment of IPAH severity and prognosis are relied solely on indicators such as N-terminal prohormone of brain natriuretic peptide, or brain natriuretic peptide, which could reflect cardiac function but fail to promptly response to the dynamic change of disease condition due to the complexity of IPAH pathogenesis. Purpose We evaluated the prognostic value of plasma proteins in addition to current clinical predictors for prognosis prediction of IPAH. Methods Olink Immune-Response panel was used to measure 92 plasma proteins in a prospective IPAH cohort with 169 patients and a mean follow-up period of 2.4 ± 0.9 years. Six machine learning methods were used to construct predictive models, including plasma proteins-based Olink models, clinical parameters-based clinical models, and integrative models. Results A total of 18 proteins showed differential expression between patients with or without adverse clinical outcomes. Three proteins (STC1, CXADR, and IL6) were strongly correlated with clinical indicators of IPAH severity and their concentrations showed upward trends along with increased risk stratification. Elevated levels of STC1, CXADR, and IL6 are all associated with an increased risk of adverse clinical events, indicated by survival analysis, restricted cubic splines, and multivariable Cox regression analysis (STC1 HR=3.474, 95%CI: 1.129-10.692, P=0.030; CXADR: HR=1.602, 95%CI: 1.074-2.387, P=0.021; IL6: HR=1.883, 95%CI: 1.376-2.576, P<0.001). The inclusion of these three proteins into the predictive model could significantly increase the predictive power on the basis of the clinical model (AUC: 0.746 vs 0.626). Conclusions High levels of STC1, CXADR, and IL6 in plasma were associated with severe clinical phenotype and increased risk of adverse clinical events in patients with IPAH. These plasma proteins could significantly improve the prognostic performance of pre-existing clinical models, facilitating the clinical management of IPAH patients.The predictive efficient of models

Plasma proteomics of all-cause and cause-specific mortality in individuals with and without type 2 diabetes: Results from two population-based KORA cohorts

Abstract Background Protein biomarkers may contribute to the identification of vulnerable subgroups for premature mortality in middle-aged and older adults. Purpose Given the distinctive impact of type 2 diabetes (T2D) on mortality rates and protein levels, this study aimed to explore the association of proteins with all-cause and cause-specific mortality separately among individuals with and without baseline T2D and to assess their impact on the prediction of all-cause mortality. Methods Using proximity extension assay technology, we measured 233 cardiovascular (CV) disease- and inflammation-related proteins in two population-based KORA (Cooperative Health Research in the Region of Augsburg) cohorts. The discovery KORA S4 study, with 15.6 years of follow-up, included 1545 participants (T2D group: 116 total, 62 CV, 31 cancer-related and 23 other-cause deaths; non-T2D group: 321 total, 114 CV, 120 cancer-related and 87 other-cause deaths). The validation KORA-Age1 study, with 6.9 years of follow-up, included 1031 participants (T2D group: 76 total, 45 CV, 19 cancer-related and 12 other-cause deaths; non-T2D group: 169 total, 74 CV, 39 cancer-related and 56 other-cause deaths). Cox regression was used to examine associations of proteins with all-cause and cause-specific mortality. Furthermore, eXtreme Gradient Boosting was applied for identification of the most relevant proteins for the prediction of all-cause mortality stratifying by baseline T2D status. The predictive performance of the protein-based model and combined model including both proteins and clinical risk factors was compared to a clinical risk factor-based model using the C-index, category-free net reclassification index (cfNRI), and relative integrated discrimination improvement (IDI). Results After validation, we identified 48 proteins associated with all-cause mortality in the T2D group and 64 in the non-T2D group, respectively (37 overlapping proteins). Out of these, 45, eight, and 32 were associated with CV, cancer-related, and other-cause mortality in the T2D group, while 53, 42, and 48 were associated with the respective cause-specific mortality outcomes in the non-T2D group in the pooled analysis of both studies. The combined model improved the prediction of all-cause mortality in both the KORA S4 and KORA-Age1 cohorts among individuals with baseline T2D (KORA-Age1: ΔC-index: 0.065, 95%=0.009-0.129; cfNRI: 0.370, 95%=0.203-0.812; IDI: 0.129, 95%=0.059-0.253) as well as those without T2D (KORA-Age1: ΔC-index: 0.037, 95%=0.015-0.071; cfNRI: 0.370, 95%=0.236-0.632; IDI: 0.053, 95%=0.028-0.104), respectively, compared to the clinical model. Conclusion This study discovered shared and unique proteins in individuals with and without T2D, unveiling the complex dynamics of mortality. Moreover, our findings support the role of proteins in improving prediction of premature mortality beyond clinical risk factors in different T2D subgroups, providing insights for future prevention strategies.

Risk prediction models for incident heart failure: a systematic review and meta-analysis

Abstract Background Heart failure (HF) risk prediction models combine multivariable patient data to estimate an individual's risk of developing HF. By detecting at-risk and early-stage patients, models may facilitate earlier intervention to prevent or delay HF development. Previous systematic reviews were unable to recommend any existing prediction models for clinical use due to insufficient evidence and lack of guidelines on appraising study quality at their time of publication. Purpose To summarize the performance of risk prediction models for incident HF and identify models for further validation and potential clinical use. Methods We searched MEDLINE and EMBASE in June 2021 for English-language studies developing or validating HF risk prediction models. Studies were also retrieved from two previous systematic reviews. We narratively summarized model characteristics (e.g. model type, predictors used, prediction horizon) and study methodology (e.g. validation methods). Performance was assessed among all models validated in ≥ 1 cohort. For all models validated in ≥ 2 cohorts, we pooled discrimination measures using random-effects meta-analyses. Calibration was descriptively summarized based on individual study results from statistical tests and graph digitization of calibration plots. Study quality was assessed using the Prediction model Risk Of Bias ASsessment Tool (PROBAST). Results Of 18,937 publications screened, 41 studies consisting of 120 prediction models were included. Twenty models were both derived and validated, 99 only derived, and 1 only validated. Risk of bias was rated as high in nearly all (94.7%) PROBAST assessments, mostly attributable to issues with analysis. Among 21 models validated in ≥ 1 cohort, most had moderate (61.9%, C-statistic 0.7 to <0.8) or high (23.8%, C-statistic 0.8 to <0.9) discrimination. In patients with low predicted risk (<10%), the calibration was adequate. Nine (42.9%) of these 21 models were presented as web-based calculators and five (23.8%) as points-based risk scores. Based on performance, number of validation cohorts, study risk of bias, and user friendliness, the Atherosclerosis Risk in Communities (ARIC), Multi-Ethnic Study of Atherosclerosis (MESA), Pooled Cohort equations to Prevent Heart Failure (PCP-HF), and Health ABC models emerged as the most promising risk scores for clinical practice. Conclusions Given their acceptable performance but high risk of bias, future studies should focus on the external validation of these models in studies of high methodological rigor. Models should be validated in a greater diversity of patient populations, particularly with respect to race. Impact analyses assessing how the clinical implementation of these models affects patient outcomes are also required prior to their routine use. Once validated, these models may help guide clinical decision making to prevent the onset of HF with early, aggressive risk factor modification.Discrimination in 21 validated modelsCharacteristics of recommended models

A Study of Recommendation Methods Based on Graph Hybrid Neural Networks and Deep Crossing

In the face of complex user behavior patterns and massive data, improving the performance of recommender system models is an urgent challenge. Traditional methods often struggle to effectively handle feature interactions and complex user-item relationships. Combining the advantages of graph neural networks and the Deep Crossing network, this paper proposes a recommendation method based on hybrid neural networks with Deep Crossing (Deep Crossing with Graph Convolution and GRU, DCGCN-GRU). First, by constructing the graph structure of users and items, higher-order feature representations are extracted, and node features are updated using a multilayer graph convolution operation. Then, the higher-order features learned by the graph convolution network are spliced and weighted with the original features to form new feature inputs. Next, a Gated Recurrent Unit (GRU) is introduced to capture the inter-feature temporal dynamic relationships and sequence information. Finally, the Deep Crossing model is utilized to learn the interactions between the fused features at multiple levels and enhance the interactions between the features. Comparative experiments on three public datasets, MovieLens-ml-25m, Book-Crossings, and Amazon Reviews’23, show that the model achieves significant improvements in accuracy, mean square error (MSE), and mean absolute error (MAE).

A Novel Natural Language Processing Model Transfer Strategy Tailored for Deep Learning Platforms

As artificial intelligence technology continues to evolve, numerous advancements enable smoother communication and collaboration between humans and computers. Natural language processing (NLP), as a key direction in the field of computer and artificial intelligence, has been attracting much attention. It focuses on the language used by people in daily life, devoting itself to improving the ability of language generation and understanding, thus making communication between people and computers more fluent and natural, effectively breaking through the problem of human-computer interaction. With the advancement of deep learning technology, the migration of NLP models onto deep learning platforms has emerged as a key trend. Model transfer, a crucial aspect of deep learning, holds significant value in enhancing the performance and efficiency of NLP models. This paper begins by outlining the fundamentals of deep learning platforms and NLP model transfer, followed by a comprehensive examination of current research progress and challenges in this field. Subsequently, we introduce a novel NLP model transfer strategy tailored for deep learning platforms and validate its effectiveness through rigorous experiments. In conclusion, the paper highlights our noteworthy advancements, pointing toward promising future developments.

Advanced Prediction Models for Scouring Around Bridge Abutments: A Comparative Study of Empirical and AI Techniques

Scouring is a major concern affecting the overall stability and safety of a bridge. The current research investigated the effectiveness of the various artificial intelligence (AI) techniques, such as artificial neural networks (ANNs), the adaptive neuro-fuzzy inference system (ANFIS), and random forest (RF), for scouring depth prediction around a bridge abutment. This study attempted to make a comparative analysis between these AI models and empirical equations developed by various researchers. The current research paper utilized a dataset of water depth (Y), flow velocity (V), discharge (Q), and sediment particle diameter (d50) from a controlled laboratory setting. An efficient optimization tool (MATLAB Optimization Tool (version R2023a)) was used to develop a scour estimation formula around bridge abutments. The findings of the current investigation demonstrated the superior performance of the AI models, especially the ANFIS model, over empirical equations by precisely capturing the non-linear and complex interactions between these parameters. Moreover, the result of the sensitivity analysis demonstrated flow velocity and discharge to be the most influencing parameters affecting the scouring depth around a bridge abutment. The results of the current research highlight the precise and accurate prediction of the scouring depth around a bridge abutment using AI models. However, the empirical equation (Equation 2) demonstrated better performance with a higher R-value of 0.90 and a lower MSE value of 0.0012 compared to other empirical equations. The findings revealed that ANFIS, when combined with neural networks and fuzzy logic systems, produced highly accurate and precise results compared to the ANN models.

A novel multi-model feature generation technique for suicide detection

Automated expert systems (AES) analyzing depression-related content on social media have piqued the interest of researchers. Depression, often linked to suicide, requires early prediction for potential life-saving interventions. In the conventional approach, psychologists conduct patient interviews or administer questionnaires to assess depression levels. However, this traditional method is plagued by limitations. Patients might not feel comfortable disclosing their true feelings to psychologists, and counselors may struggle to accurately predict situations due to limited data. In this context, social media emerges as a potentially valuable resource. Given the widespread use of social media in daily life, individuals often express their nature and mental state through their online posts. AES can efficiently analyze vast amounts of social media content to predict depression levels in individuals at an early stage. This study contributes to this endeavor by proposing an innovative approach for predicting suicide risks using social media content and machine learning techniques. A novel multi-model feature generation technique is employed to enhance the performance of machine learning models. This technique involves the use of a feature extraction method known as term frequency-inverse document frequency (TF-IDF), combined with two machine learning models: logistic regression (LR) and support vector machine (SVM). The proposed technique calculates probabilities for each sample in the dataset, resulting in a new feature set referred to as the probability-based feature set (ProBFS). This ProBFS is compact yet highly correlated with the target classes in the dataset. The utilization of concise and correlated features yields significant outcomes. The SVM model achieves an impressive accuracy score of 0.96 using ProBFS while maintaining a low computational time of 5.63 seconds even when dealing with extensive datasets. Furthermore, a comparison with state-of-the-art approaches is conducted to demonstrate the significance of the proposed method.

Importance and limits of cardiovascular risk factors on the prediction of SCD using machine learning approach

Abstract Background SCD is mainly due to lethal ventricular arrhythmias and occurs often among patients with ischemic heart disease. SCD and ischemic heart disease share a similar pattern of cardiovascular risk factors (CVRF). Commonly developed risk scores have been focused on cardiac-related risk factors. We aimed to investigate whether non-CVRF variables could enhance predictive performance beyond standard CVRF of SCD. Method We compared 2 different strategies for variable inclusion. First, we built a prediction model containing EHR data restricted to cardiovascular diseases and risk factors (CVD MR model) that occurred up to 5 years before SCD. We selected all available main cardiovascular variables as surrogate markers for coronary artery disease, stroke, diabetes, hypertension, smoking status, obesity, lipid disorders and chronic renal failure. On the other hand, our global approach (ALL MR model) includes all medical records codes, without any prior selection. To estimate the risk of SCD over three months, we trained a machine learning model on EHR data representing 8,566,229 drug prescriptions and 801,352 hospital diagnoses up to five years prior to SCD. The data were obtained from a cohort of 12,338 SCD in France and 12,338 controls from 2011 to 2015. We then validated the results on two external cohorts: one temporal in the same area between 2016 and 2020 with 11,620 SCD and 11,620 controls and one geographical from the USA with 892 SCD and 892 controls from 2013 to 2021. Results The CVD MR + ML model (CatBoost algorithm) yielded moderate performances with an AUC of 0.68 (95%CI 0.65-0.68), a sensitivity of 45%.(95%CI 42-47) and a specificity of 82% (95%CI 80-84). We found that the EHR model with all medical records (All MR + ML model) offered better performances. In the derivation cohort, this model achieved an AUC of 0.80 (95% CI: 0.78–0.82) with a sensitivity of 67% (95% CI 64-69), a specificity of 80% (95% CI 78-82) (Figure 1). The logistic regression (All MR + LR) applied on all medical records performed better than the CatBoost algorithm when restricted to CVD medical records (Figure 1). Conclusion The inclusion of all potential variables beyond the usual CVRF significantly improves the performances of SCD prediction models, independently of the methods (AI and logistic regression). Figure 1 : Receiver operating characteristic curves of the different models (ALL MR + ML, ALL MR + LR, CVD MR + ML and CVD MR + LR) in the derivation cohort and two validation cohorts (temporal and geographical)Figure 1.ROC curves

A comparative ensemble approach to bedload prediction using metaheuristic machine learning

An adequate bedload prediction is a challenging task in hydraulic engineering because of the complex sediment transport processes and corresponding environmental factors. The current study introduces a novel comparative ensemble approach using metaheuristic machine learning (ML) models to enhance the accuracy of bedload prediction using data from laboratory flume experiments. To uncover key insights in bedload transport prediction, several models are employed, such as K-Nearest Neighbours (KNN), Extra Trees Regressor (ETR), Linear Regression (LR), Random Forest (RF), Bagging Regressor (BR), and XGBoost (XGB). The coefficient of determination (R2) and root mean square error (RMSE) values vary between various models; however, XGB showed R2 = 0.99 and RMSE = 0.11. The sensitivity analysis emphasizes the crucial role of the Shields parameter in bedload prediction, while the SHAP analysis highlights the substantial influence of the XGB model in enhancing predictive accuracy. The REC curves show that BR, XGB, and RF, outperformed KNN and LR models. Furthermore, a graphical user interface has also been developed to facilitate user interaction with the predictive models, allowing for easier visualization, analysis, and interpretation of bedload transport predictions. Additionally, k-fold cross-validation was performed to assess the performance, consistency and robustness of the ML models. Thus, it can be concluded from the current study that the utilization of ML algorithms can improve accuracy, providing valuable insights for hydraulic engineers and highlighting the importance of ML models in civil engineering practices, particularly in bedload transport prediction.