|
|
Atrial fibrillation detection using time-domain methods
Sámel, Maroš ; Ronzhina, Marina (referee) ; Janoušek, Oto (advisor)
The purpose of this work is to get better understanding of the problems of atrial fibrillation followed by the processing methods for detection of atrial fibrillation based on an analysis of the ventricular rhythm. Methods that are described: method of medians, methods based on histograms (using coefficient of variation and density histograms of RR and RR intervals), method using RdR map, method using complexity of RR intervals and methods based on a statistical framework by Gaussian distribution and Laplace function The practical part of this work is focused on comparing these methods and selecting the best one – method of medians is the chosen one. This is followed by implementation of this method in the Matlab and testing it on the real data from a selected database. In the end, the detection ability of our method is evaluated and compared with the thoeretical detection abilities of the other methods. Our algorithm for detection of atrial fibrillations, based on the median method, achieved excellent results with the highest value of specificity obtained at 93,976%, sensitivity at 89,182% and AUC (area under the curve) at 0,973.
|
|
|
Atrial fibrillation detection using time-domain methods
Sámel, Maroš ; Ronzhina, Marina (referee) ; Janoušek, Oto (advisor)
The purpose of this work is to get better understanding of the problems of atrial fibrillation followed by the processing methods for detection of atrial fibrillation based on an analysis of the ventricular rhythm. Methods that are described: method of medians, methods based on histograms (using coefficient of variation and density histograms of RR and RR intervals), method using RdR map, method using complexity of RR intervals and methods based on a statistical framework by Gaussian distribution and Laplace function The practical part of this work is focused on comparing these methods and selecting the best one – method of medians is the chosen one. This is followed by implementation of this method in the Matlab and testing it on the real data from a selected database. In the end, the detection ability of our method is evaluated and compared with the thoeretical detection abilities of the other methods. Our algorithm for detection of atrial fibrillations, based on the median method, achieved excellent results with the highest value of specificity obtained at 93,976%, sensitivity at 89,182% and AUC (area under the curve) at 0,973.
|
guest ::
