Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Introduction: Heart rate variability (HRV) reflects the activity of the autonomic nervous system on the heart rate. Reduced HRV has been associated with adverse outcomes. Endurance athletes show a better resting HRV. However, it is unknown whether starting endurance training at an older age results in the same beneficial cardiac autonomic changes similar to lifelong endurance training. Methods: We conducted a single centre sub-analysis of the Master@Heart study, comparing time and frequency domain HRV components measured with 24h Holter monitoring, between male senior athletes with lifelong endurance training (n = 60), senior athletes with late-onset endurance training (n = 56) and healthy non-athletic subjects (n = 59). Results: 60 lifelong endurance athletes (age: median 55 years, range 25 years), 56 late-onset endurance athletes (age: median 56 years, range 26 years) and 59 healthy non-athletes (age: median 55 years, range 25 years) were included. Resting average heart rate was significantly lower in both athlete groups (63 ± 9 and 61 ± 7 bpm) compared to healthy controls (68 ± 9 bpm, p < 0.001). In the time domain RMSSD was 32% higher in late onset athletes and 22% higher in lifelong athletes (p = 0.013). SDNN (p < 0.001), SDANN (p < 0.001) and pNN50 (p = 0.029) were significantly higher in the athlete groups as well as in the frequency domain total power (p < 0.001) and VLF power (p = 0.001) compared to healthy controls. Conclusion: Both late-onset and lifelong endurance athletes have higher parasympathetic modulation and higher overall resting HRV compared with healthy controls. Thus, endurance training improved cardiac autonomic function even if the exercise was initiated later in life, which may reduce cardiovascular risk and contribute to healthy ageing.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Sudden Cardiac Death is one of the most common causes of mortality within cardiovascular diseases, often resulting from arrhythmic arrests. An important example of a preventive measure against these arrhythmias is the ICD (Implantable Cardioverter Defibrillator), a device able to deliver shocks to revert these life-threatening arrhythmias. Despite the effectiveness of these devices, the identification of the patients who stand to gain the most from implantation is a challenging task. By improving risk stratification methods, resources can be allocated to the right patients, maximizing life-saving interventions, without wasting valuable assets. This master thesis applies two deep learning methods on electrocardiographic data and adapts their models to be compatible with three-dimensional vectorcardiographic data. By applying this approach as time series analysis, this thesis aspires to augment the precision in identifying patients who are at an elevated risk. This thesis utilized a dataset obtained from the University Hospitals of Leuven: it contains 1242 ECG recordings of patients equipped with an ICD. The method consisted of two key phases: an initial classification task focused on understanding the basics of ECG (Electrocardiogram) and VCG (Vectorcardiogram) and its use in deep learning, and a second part focusing on survival analysis. In the exploratory classification of patients suffering from ICM (Ischemic Cardiomyopathy) and DCM (Dilated Cardiomyopathy), both deep learning techniques yielded similar results, achieving an accuracy of approximately 71%. For the survival analysis, the results revealed that using risk scores produced by the neural networks-representing the chances of the event occurring-is not sufficient on its own. However, when these risk scores were included as a feature in a Cox Regression, an improved concordance index was observed across all endpoints. This indicates that incorporating these risk scores can enhance the predictive accuracy of the risk stratification models, potentially leading to more effective ICD management. The findings underscore the potential of deep learning techniques applied to vectorcardiographic data. While the output of the deep learning methods has room for growth, they highlight the need for further research. Refining the models, enhancing datasets, and addressing challenges in data preprocessing will lead to better results in future studies.