Abstract

Atrial fibrillation (AF) is the most common cardiac arrhythmia and is the cause of 15% of all ischemic strokes in the United States. In many cases, AF is episodic and/or asymptomatic and as a result, there is a strong need for algorithms capable of quickly and reliably detecting AF, even in cases where the heart rate is controlled through medication or with a pacemaker. Current RR interval (RRI)-based algorithms do not directly target atrial activity, cannot detect AF when the heart rate is controlled, and analyze relatively long intervals of the electrocardiography (ECG) to make an AF determination. This work proposes an algorithm for patient-specific, single-beat, rate-independent AF identification based on atrial activity (AA) analysis. The proposed algorithm develops a statistical model to describe the distribution of features extracted from AA during normal sinus rhythm (NSR). First, ECG segments preceding QRS complexes are identified potential P waves. A total of nine features - three higher order statistics (HOS) features and six features obtained through downsampling - are extracted from the P wave segment under consideration. The Expectation-Maximization algorithm is applied to a training set to create a multivariate Gaussian Mixture Model (GMM) of the feature space. This model is used to identify P wave absence (PWA) and, in turn, AF. An optional post-processing stage which takes a majority vote of successive outputs is applied to improve classifier performance. To evaluate the performance of the classifier, the algorithm was tested on 20 records in the MIT-BIH Atrial Fibrillation Database. Single-beat classification showed a sensitivity(Se) of 91.98%, a specificity(Sp) of 86.18%, a positive predictive value(PPV) of 70.70% and an error rate(Err) of 13.02%. Classification combining seven beats showed a Se of 99.28%, a Sp of 90.21%, a PPV of 80.42% and an Err of 7.12%. The presented algorithm has a classification performance comparable to current RRI-based algorithms yet is rate-independent and capable of making an AF determination in a single beat.

Library of Congress Subject Headings

Atrial fibrillation--Data processing; Electrocardiography--Data processing

Publication Date

1-2014

Document Type

Thesis

Student Type

Graduate

Degree Name

Electrical Engineering (MS)

Department, Program, or Center

Electrical Engineering (KGCOE)

Advisor

Behnaz Ghoraani

Advisor/Committee Member

Daniel Phillips

Advisor/Committee Member

Sohail Dianat

Comments

Physical copy available from RIT's Wallace Library at RC685.A72 L34 2014

Campus

RIT – Main Campus

Plan Codes

EEEE-MS

Download

Share

COinS