Atrial fibrillation detection in outpatient electrocardiogram monitoring: An algorithmic crowdsourcing approach

Background Atrial fibrillation (AFib) is the most common cardiac arrhythmia associated with stroke, blood clots, heart failure, coronary artery disease, and/or death. Multiple methods have been proposed for AFib detection, with varying performances, but no single approach appears to be optimal. We hypothesized that each state-of-the-art algorithm is appropriate for different subsets of patients and provides some independent information. Therefore, a set of suitably chosen algorithms, combined in a weighted voting framework, will provide a superior performance to any single algorithm. Methods We investigate and modify 38 state-of-the-art AFib classification algorithms for a single-lead ambulatory electrocardiogram (ECG) monitoring device. All algorithms are ranked using a random forest classifier and an expert-labeled training dataset of 2,532 recordings. The seven top-ranked algorithms are combined by using an optimized weighting approach. Results The proposed fusion algorithm, when validated on a separate test dataset consisting of 4,644 recordings, resulted in an area under the receiver operating characteristic (ROC) curve of 0.99. The sensitivity, specificity, positive-predictive-value (PPV), negative-predictive-value (NPV), and F1-score of the proposed algorithm were 0.93, 0.97, 0.87, 0.99, and 0.90, respectively, which were all superior to any single algorithm or any previously published. Conclusion This study demonstrates how a set of well-chosen independent algorithms and a voting mechanism to fuse the outputs of the algorithms, outperforms any single state-of-the-art algorithm for AFib detection. The proposed framework is a case study for the general notion of crowdsourcing between open-source algorithms in healthcare applications. The extension of this framework to similar applications may significantly save time, effort, and resources, by combining readily existing algorithms. It is also a step toward the democratization of artificial intelligence and its application in healthcare.

Download Full-text

Usefulness of the ACTEL Score to Predict Atrial Fibrillation in Patients with Cryptogenic Stroke

Cardiology ◽

10.1159/000505262 ◽

2020 ◽

Vol 145 (3) ◽

pp. 168-177 ◽

Cited By ~ 1

Author(s):

Antonio Muscari ◽

Pietro Barone ◽

Luca Faccioli ◽

Marco Ghinelli ◽

Marco Pastore Trossello ◽

...

Keyword(s):

Atrial Fibrillation ◽

Ischemic Stroke ◽

Positive Predictive Value ◽

Predictive Value ◽

Characteristic Curve ◽

Cryptogenic Stroke ◽

Left Ventricular ◽

Control Group ◽

End Diastolic Volume ◽

Electrocardiogram Monitoring

Introduction: To assess the probability of undetected atrial fibrillation (AF) in patients with ischemic stroke, we previously compared patients who were first diagnosed with AF with patients with large or small artery disease and obtained the MrWALLETS 8-item scoring system. In the present study, we utilized cryptogenic strokes (CS) as the control group, as AF is normally sought among CS patients. Methods: We retrospectively examined 191 ischemic stroke patients (72.5 ± 12.6 years), 68 with first diagnosed AF and 123 with CS, who had undergone 2 brain CT scans, echocardiography, carotid/vertebral ultrasound, continuous electrocardiogram monitoring and anamnestic/laboratory search for cardiovascular risk factors. Results: In logistic regression, 5 variables were independently associated with AF, forming the “ACTEL” score: Age ≥75 years (OR 2.42, 95% CI 1.18–4.96, p = 0.02; +1 point); hyperCholesterolemia (OR 0.38, 95% CI 0.18–0.78, p = 0.009; –1 point); Tricuspid regurgitation ≥ mild-to-moderate (OR 4.99, 95% CI 1.63–15.27, p = 0.005; +1 point); left ventricular End-diastolic volume <65 mL (OR 7.43, 95% CI 2.44–22.6, p = 0.0004; +1 point); Left atrium ≥4 cm (OR 4.57, 95% CI 1.97–10.62, p = 0.0004; +1 point). The algebraic sum of these points may range from –1 to +4. For AF identification, the area under the receiver operating characteristic curve was 0.80 (95% CI 0.73–0.87). With a cutoff of ≥2, positive predictive value was 80.8%, specificity 92.7% and sensitivity 55.9%. Conclusions: The ACTEL score, a simplified and improved version of the MrWALLETS score, allows the identification of patients with first diagnosed AF, in the context of CSs, with a high positive predictive value.

Download Full-text

Correlation of magnetic resonance imaging and post-ablation endoscopy to detect oesophageal thermal injury in patients after atrial fibrillation ablation: MRI-EDEL-study

EP Europace ◽

10.1093/europace/euaa081 ◽

2020 ◽

Vol 22 (7) ◽

pp. 1009-1016

Author(s):

Philipp Halbfass ◽

Lukas Lehmkuhl ◽

Borek Foldyna ◽

Artur Berkovitz ◽

Kai Sonne ◽

...

Keyword(s):

Magnetic Resonance Imaging ◽

Atrial Fibrillation ◽

Magnetic Resonance ◽

Predictive Value ◽

Diagnostic Value ◽

Resonance Imaging ◽

Oesophageal Injury ◽

Mri Sequences ◽

Sensitivity Specificity ◽

Signal Intensities

Abstract Aims To correlate oesophageal magnetic resonance imaging (MRI) abnormalities with ablation-induced oesophageal injury detected in endoscopy. Methods and results Ablation-naïve patients with atrial fibrillation (AF), who underwent ablation using a contact force sensing irrigated radiofrequency ablation catheter, received a cardiac MRI on the day of ablation, and post-ablation oesophageal endoscopy (OE) 1 day after ablation. Two MRI expert readers recorded presence of abnormal oesophageal tissue signal intensities, defined as increased oesophageal signal in T2-fat-saturated (T2fs), short-tau inversion-recovery (STIR), or late gadolinium enhancement (LGE) sequences. Oesophageal endoscopy was performed by experienced operators. Finally, we correlated the presence of any affection with endoscopically detected oesophageal thermal lesions (EDEL). Among 50 consecutive patients (age 67 ± 7 years, 60% male), who received post-ablation MRI and OE, complete MRI data were available in 44 of 50 (88%) patients. In OE, 7 of 50 (14%) presented with EDEL (Category 1 lesion: erosion n = 3, Category 2 lesion: ulcer n = 4). Among those with EDEL, 6 of 7 (86%) patients presented with increased signal intensities in all three MRI sequences, while only 2 of 37 (5%) showed hyperintensities in all three MRI sequences and negative endoscopy. Correspondingly, sensitivity, specificity, positive predictive value, and negative predictive value (NPV) for MRI (increased signal in T2fs, STIR, and LGE) were 86%, 95%, 75%, and 97%, respectively. Conclusion Increased signal intensity in T2fs, STIR, and LGE represents independent markers of EDEL. In particular, the combination of all three has the highest diagnostic value. Hence, MRI may represent an accurate, non-invasive method to exclude acute oesophageal injury after AF ablation (NPV: 97%).

Download Full-text

DNC4mC-Deep: Identification and Analysis of DNA N4-Methylcytosine Sites Based on Different Encoding Schemes By Using Deep Learning

Cells ◽

10.3390/cells9081756 ◽

2020 ◽

Vol 9 (8) ◽

pp. 1756 ◽

Cited By ~ 4

Author(s):

Abdul Wahab ◽

Omid Mahmoudi ◽

Jeehong Kim ◽

Kil To Chong

Keyword(s):

Deep Learning ◽

Protein Interactions ◽

State Of The Art ◽

Critical Role ◽

Regulation Of Gene Expression ◽

The State ◽

Superior Performance ◽

Training Dataset ◽

Conformation Stability ◽

Deep Learning Model

N4-methylcytosine as one kind of modification of DNA has a critical role which alters genetic performance such as protein interactions, conformation, stability in DNA as well as the regulation of gene expression same cell developmental and genomic imprinting. Some different 4mC site identifiers have been proposed for various species. Herein, we proposed a computational model, DNC4mC-Deep, including six encoding techniques plus a deep learning model to predict 4mC sites in the genome of F. vesca, R. chinensis, and Cross-species dataset. It was demonstrated by the 10-fold cross-validation test to get superior performance. The DNC4mC-Deep obtained 0.829 and 0.929 of MCC on F. vesca and R. chinensis training dataset, respectively, and 0.814 on cross-species. This means the proposed method outperforms the state-of-the-art predictors at least 0.284 and 0.265 on F. vesca and R. chinensis training dataset in turn. Furthermore, the DNC4mC-Deep achieved 0.635 and 0.565 of MCC on F. vesca and R. chinensis independent dataset, respectively, and 0.562 on cross-species which shows it can achieve the best performance to predict 4mC sites as compared to the state-of-the-art predictor.

Download Full-text

A Urine-Based Liquid Biopsy Method for Detection of Upper Tract Urinary Carcinoma

Frontiers in Oncology ◽

10.3389/fonc.2020.597486 ◽

2021 ◽

Vol 10 ◽

Author(s):

Yansheng Xu ◽

Xin Ma ◽

Xing Ai ◽

Jiangping Gao ◽

Yiming Liang ◽

...

Keyword(s):

Logistic Regression ◽

Negative Predictive Value ◽

High Throughput ◽

Predictive Value ◽

Liquid Biopsy ◽

High Throughput Sequencing ◽

Superior Performance ◽

Detection Methods ◽

Training Dataset ◽

Upper Tract

BackgroundConventional clinical detection methods such as CT, urine cytology, and ureteroscopy display low sensitivity and/or are invasive in the diagnosis of upper tract urinary carcinoma (UTUC), a factor precluding their use. Previous studies on urine biopsy have not shown satisfactory sensitivity and specificity in the application of both gene mutation or gene methylation panels. Therefore, these unfavorable factors call for an urgent need for a sensitive and non-invasive method for the diagnosis of UTUC.MethodsIn this study, a total of 161 hematuria patients were enrolled with (n = 69) or without (n = 92) UTUC. High-throughput sequencing of 17 genes and methylation analysis for ONECUT2 CpG sites were combined as a liquid biopsy test panel. Further, a logistic regression prediction model that contained several significant features was used to evaluate the risk of UTUC in these patients.ResultsIn total, 86 UTUC− and 64 UTUC+ case samples were enrolled for the analysis. A logistic regression analysis of significant features including age, the mutation status of TERT promoter, and ONECUT2 methylation level resulted in an optimal model with a sensitivity of 94.0%, a specificity of 93.1%, the positive predictive value of 92.2% and a negative predictive value of 94.7%. Notably, the area under the curve (AUC) was 0.957 in the training dataset while internal validation produced an AUC of 0.962. It is worth noting that during follow-up, a patient diagnosed with ureteral inflammation at the time of diagnosis exhibiting both positive mutation and methylation test results was diagnosed with ureteral carcinoma 17 months after his enrollment.ConclusionThis work utilized the epigenetic biomarker ONECUT2 for the first time in the detection of UTUC and discovered its superior performance. To improve its sensitivity, we combined the biomarker with high-throughput sequencing of 17 genes test. It was found that the selected logistic regression model diagnosed with ureteral cancer can evaluate upper tract urinary carcinoma risk of patients with hematuria and outperform other existing panels in providing clinical recommendations for the diagnosis of UTUC. Moreover, its high negative predictive value is conducive to rule to exclude patients without UTUC.

Download Full-text

Use of Simplified HAS-BLED Score for Predicting Bleeding Events in Anticoagulated Patients with Atrial Fibrillation

Journal of the Medical Association of Thailand ◽

10.35755/jmedassocthai.2021.05.12142 ◽

2021 ◽

Vol 104 (5) ◽

pp. 802-806

Keyword(s):

Atrial Fibrillation ◽

Risk Prediction ◽

Major Bleeding ◽

Predictive Value ◽

Statistical Significance ◽

Bleeding Risk ◽

Minor Bleeding ◽

Bleeding Events ◽

Sensitivity Specificity ◽

Anticoagulated Patients

Objective: To demonstrate bleeding risk prediction of simplified HAS-BLED (sHAS-BLED) score in anticoagulated patients with atrial fibrillation (AF). Materials and Methods: AF patients receiving warfarin were retrospectively recruited in Central Chest Institute of Thailand between October 2012 and December 2017. The main outcome was total bleeding including major bleeding, clinically relevant non-major bleeding or minor bleeding. The chi-square test or Fisher’s exact test was used to compare the main outcome between sHAS-BLED and conventional HAS-BLED (cHAS-BLED) scores. A sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of sHAS-BLED were calculated. The discrimination performances of sHAS-BLED and cHAS-BLED scores were demonstrated with c-statistics. Results: One hundred ten patients were recruited. The mean age was 70.53±9.58 years. The average sHAS-BLED and cHAS-BLED scores were 2.23±0.79 and 1.95±0.83, respectively. The patients with sHAS-BLED score of 3 or more had 15 total bleeding events (37.50%) while those with score of less than 3 had 13 total bleeding events (18.57%). Those with sHAS-BLED score of 3 or more had more total bleeding than those with score of less than 3 with statistical significance (odds ratio 2.63; 95% CI 1.09 to 6.25; p=0.049). A sensitivity, specificity, PPV, and NPV of sHAS-BLED score were 53.57%, 69.51%, 37.50%, and 81.43%, respectively. The discrimination performances of sHAS-BLED and cHAS-BLED scores were demonstrated with c-statistics of 0.65 and 0.67, respectively. Conclusion: The sHAS-BLED score can be used for bleeding risk prediction in anticoagulated AF patients compared with cHAS-BLED score. Keywords: Simplified HAS-BLED, Atrial fibrillation, Anticoagulant, Bleeding, SAMe-TT₂R₂

Download Full-text

A Urine-Based Liquid Biopsy Method for Detection of Upper Tract Urinary Carcinoma

10.21203/rs.3.rs-53500/v1 ◽

2020 ◽

Author(s):

Yansheng Xu ◽

Xin Ma ◽

Xing Ai ◽

Jiangping Gao ◽

Yiming Liang ◽

...

Keyword(s):

Logistic Regression ◽

Negative Predictive Value ◽

High Throughput ◽

Predictive Value ◽

Liquid Biopsy ◽

High Throughput Sequencing ◽

Superior Performance ◽

Detection Methods ◽

Training Dataset ◽

Upper Tract

Abstract Background: Conventional clinical detection methods such as CT, urine cytology, and ureteroscopy display low sensitivity and/or are invasive in the diagnosis of upper tract urinary carcinoma (UTUC), a factor precluding their use. Previous studies on urine biopsy have not shown satisfactory sensitivity and specificity in the application of both gene mutation or gene methylation panels. Therefore, these unfavorable factors call for an urgent need for a sensitive and non-invasive method for the diagnosis of UTUC.Methods: In this study, a total of 161 hematuria patients were enrolled with (n=69) or without (n=92) UTUC. High-throughput sequencing of 17 genes and methylation analysis for ONECUT2 CpG sites were combined as a liquid biopsy test panel. Further, a logistic regression prediction model that contained several significant features was used to evaluate the risk of UTUC in these patients.Results: In total, 86 UTUC- and 64 UTUC+ case samples were enrolled for the analysis. A logistic regression analysis of significant features including age, the mutation status of TERT promoter and ONECUT2 methylation level resulted inan optimal model with a sensitivity of 94.0%, a specificity of 93.1%, the positive predictive value of 92.2% and a negative predictive value of 94.7%. Notably, the area under the curve (AUC) was 0.957 in the training dataset while internal validation produced an AUC of 0.962. It is worth noting that during follow-up, a patient diagnosed with ureteral inflammation at the time of diagnosis exhibiting both positive mutation and methylation test results was diagnosed with ureteral carcinoma 17 months after his enrollment.Conclusion: This work utilized the epigenetic biomarker ONECUT2 for the first time in the detection of UTUC and discovered its superior performance. To improve its sensitivity, we combined the biomarker with a high-throughput sequencing of 17 genes test. It was found that the selected logistic regression model diagnosed with ureteral cancerThetcan evaluate upper tract urinary carcinoma risk of patients with hematuria and outperform other existing panels in providing clinical recommendations for the diagnosis of UTUC. Moreover, its high negative predictive value is conducive to rule to exclude patients without UTUC.

Download Full-text

A Clinical Study on Atrial Fibrillation, Premature Ventricular Contraction, and Premature Atrial Contraction Screening Based on an ECG Deep Learning Model (Preprint)

10.2196/preprints.29734 ◽

2021 ◽

Author(s):

Jianyuan Hong ◽

Hua-Jung Li ◽

Chung-Chi Yang ◽

Chi-Lu Han ◽

Jui-Chien Hsieh

Keyword(s):

Atrial Fibrillation ◽

Deep Learning ◽

Premature Ventricular Contraction ◽

Learning Model ◽

Training Dataset ◽

Ventricular Contraction ◽

Atrial Contraction ◽

Premature Atrial Contraction ◽

Sensitivity Specificity ◽

Deep Learning Model

BACKGROUND As the results of this study indicate, electrocardiography (ECG) devices generating interpretations of atrial fibrillation (AF), premature ventricular contraction (PVC), and premature atrial contraction (PAC) have high ratios of false-positive errors. OBJECTIVE The aim of this study was to develop an electrocardiogram (ECG) interpreter to improve the performance of AF, PVC, and PAC screening based on an ECG. METHODS In this study, we first adopted a deep learning model to delineate ECG features such as P, QRS, and T waves based on 1160 8–10-s lead I or lead II ECG signals whose ECG device interpretation is AF as a training dataset. Second, a sliding window with 3-RR intervals in length is applied to the raw ECG to examine the delineated features in the window, and the ECG interpretation is then determined based on experiences of cardiologists. RESULTS The results indicate the following: (1) This delineator achieves a good performance on P-, QRS-, and T- wave delineation with a sensitivity/specificity of 0.94/0.98, 1.00/0.99, and 0.97/0.98, respectively, in 48 10-s test ECGs mixed with AF and non-AF ECGs. (2) As compared to ECG-device generated interpretations, the precision of the detection of AF, PVC, and PAC in this study was increased from 0.77 to 0.86, 0.76 to 0.84, and 0.82 to 0.87 in 188 10-s test ECGs. Finally, (3) the F1 scores on the detection of AF, PVC, and PAC were 0.92, 0.91, and 0.83, respectively. CONCLUSIONS In conclusion, this study improved the accuracy of ECG device interpretations, and we believe that the results can bridge the gap between research and clinical practice.

Download Full-text

External continuous ECG versus loop recording for atrial fibrillation detection in patients who had a stroke

Heart ◽

10.1136/heartjnl-2018-314186 ◽

2019 ◽

Vol 105 (11) ◽

pp. 848-854 ◽

Cited By ~ 5

Author(s):

Michala Herskind Sejr ◽

Ole May ◽

Dorte Damgaard ◽

Birgitte Forsom Sandal ◽

Jens Cosedis Nielsen

Keyword(s):

Atrial Fibrillation ◽

Positive Predictive Value ◽

Predictive Value ◽

Gold Standard ◽

Primary Outcome ◽

External Loop ◽

Registration Number ◽

Ischaemic Attack ◽

Sensitivity Specificity ◽

Rule Out

BackgroundDetection of atrial fibrillation (AF) in patients who had ischaemic stroke and transient ischaemic attack (IS/TIA) is recommended. We aimed to compare external loop recording (ELR) against simultaneous continuous ECG recording for AF detection in patients who had acute IS/TIA and determine sensitivity, specificity and positive predictive value of AF detection using ELR. We hypothesised ELR to detect 15% fewer patients with AF than continuous ECG recording.MethodsIn this prospective cohort study, we included 1412 patients who had acute IS/TIA without prior AF. Monitoring was 48 hours. Primary outcome was AF >30 s. Cardiologist verified AF in continuous ECG was gold standard.ResultsIn continuous ECG, 38 (2.7%) patients had AF. ELR automatically categorised 219/1412 patients (15.5%) with AF, including 32/38 (85%) patients with AF in continuous ECG. After cardiologist adjudication of ELR recordings, AF was diagnosed in 57/219 patients, of which 32 (56%) had AF in continuous ECG. For adjudicated AF detection by ELR, sensitivity was 84%, 95% CI (69% to 94%), specificity was 98%, 95% CI (97% to 99%) and positive predictive value was 56%, 95% CI (42% to 69%).ConclusionAutomatic AF detection with ELR results in an AF diagnosis in more than five patients without AF for each patient with AF as verified in continuous ECG. For adjudicated AF detection by ELR, sensitivity was confirmed to 84% and specificity 98%. Automatic ELR as investigated in this study may be considered to rule out AF, but it is not suitable as a single monitoring device for AF screening in patients early after stroke.Trial registration numberNCT02155907.

Download Full-text

Ensemble of Machine-Learning Methods for Predicting Gully Erosion Susceptibility

Remote Sensing ◽

10.3390/rs12223675 ◽

2020 ◽

Vol 12 (22) ◽

pp. 3675

Author(s):

Subodh Chandra Pal ◽

Alireza Arabameri ◽

Thomas Blaschke ◽

Indrajit Chowdhuri ◽

Asish Saha ◽

...

Keyword(s):

Machine Learning ◽

Land Degradation ◽

Predictive Value ◽

Research Study ◽

Research Work ◽

Gully Erosion ◽

Training Dataset ◽

Operating Characteristics ◽

Boosted Regression Tree ◽

Sensitivity Specificity

Gully formation through water-induced soil erosion and related to devastating land degradation is often a quasi-normal threat to human life, as it is responsible for huge loss of surface soil. Therefore, gully erosion susceptibility (GES) mapping is necessary in order to reduce the adverse effect of land degradation and diminishes this type of harmful consequences. The principle goal of the present research study is to develop GES maps for the Garhbeta I Community Development (C.D.) Block; West Bengal, India, by using a machine learning algorithm (MLA) of boosted regression tree (BRT), bagging and the ensemble of BRT-bagging with K-fold cross validation (CV) resampling techniques. The combination of the aforementioned MLAs with resampling approaches is state-of-the-art soft computing, not often used in GES evaluation. In further progress of our research work, here we used a total of 20 gully erosion conditioning factors (GECFs) and a total of 199 gully head cut points for modelling GES. The variables’ importance, which is responsible for gully erosion, was determined based on the random forest (RF) algorithm among the several GECFs used in this study. The output result of the model’s performance was validated through a receiver operating characteristics-area under curve (ROC-AUC), sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) statistical analysis. The predicted result shows that the ensemble of BRT-bagging is the most well fitted for GES where AUC value in K-3 fold is 0.972, whereas the value of AUC in sensitivity, specificity, PPV and NPV is 0.94, 0.93, 0.96 and 0.93, respectively, in a training dataset, and followed by the bagging and BRT model. Thus, from the predictive performance of this research study it is concluded that the ensemble of BRT-Bagging can be applied as a new approach for further studies in spatial prediction of GES. The outcome of this work can be helpful to policy makers in implementing remedial measures to minimize damages caused by gully erosion.

Download Full-text

SAMe-TT₂R₂ Score for Prediction of Anticoagulation Control in Thai Patients with Non-Valvular Atrial Fibrillation

Journal of the Medical Association of Thailand ◽

10.35755/jmedassocthai.2021.06.12415 ◽

2021 ◽

Vol 104 (6) ◽

pp. 959-963

Keyword(s):

Atrial Fibrillation ◽

Positive Predictive Value ◽

Sensitivity And Specificity ◽

Predictive Value ◽

International Normalized Ratio ◽

Control Objective ◽

Anticoagulation Control ◽

Sensitivity Specificity

Background: The quality of anticoagulation control is an important determination of thromboembolism and bleeding in patients with non-valvular atrial fibrillation. Previous trials have shown that SAMe-TT₂R₂ score could be used for prediction of anticoagulation control. Objective: To predict labile international normalized ratio (INR) by SAMe-TT₂R₂ score in Thai patients with non-valvular atrial fibrillation. Materials and Methods: The author retrospectively studied patients with non-valvularatrial fibrillation at Pranangklao Hospital between January 2019 and October 2020. Results: One hundred thirty patients were enrolled. The average ages of the patients were 67.5±10.2 years. The average SAMe-TT₂R₂ scores were 3.2±0.8 and the average CHA₂DS₂-VASc score was 3.3±1.4. Most patients had hypertension and dyslipidemia. Most patients were prescribed betablockers. Most patients had time in therapeutic range (TTR) lower than 65. The present study has shown that patients with SAMe-TT₂R₂ score of 3 or more has also had high proportion of labile INR with statical significance. The sensitivity, specificity, positive predictive value, and negative predictive value of different cut-offs of SAMe-TT₂R₂ score greater than 2 and SAMe-TT₂R₂ score when excluding race showed improvement of the sensitivity and specificity for prediction of labile INR. Conclusion: Labile INR was predicted by SAMe-TT₂R₂ score and the sensitivity and specificity increased in SAMe-TT₂R₂ score when excluding race. Keywords: SAMe-TT₂R₂ score; Non-valvular atrial fibrillation; Anticoagulation control

Download Full-text