Automated Detection of Premature Ventricular Contraction Using Recurrence Quantification Analysis on Heart Rate Signals

作者：Hock Tay Swee^*; Faust Oliver; Lim Teik Cheng; Yu Wenwei

来源：Journal of Medical Imaging and Health Informatics, 2013, 3(3): 462-469.

DOI：10.1166/jmihi.2013.1181

• Diagnosis Support System
• Graphical User Interface

Heart Rate (HR) signals are important measures which can be used to assess the cardiac health of a patient. The signal even provides indicators which foreshadow impeding cardiac diseases, such as Preventricular Contraction (PVC). These indicators can be present at all times or they may occur at random. Without automated support, doctors have to spend lots of time studying and pinpointing abnormalities which indicate a heart. condition. To assist in the diagnosis, we propose a Computer-Aided Diagnosis (CAD) which differentiates between HR signals from normal subjects and PVC patients. We have used the nonlinear method of Recurrence Quantification Analysis (RQA) to extract 12 features from Recurrence Plot (RPs). Out of these 12 features, only the eight most significant features were fed to three classifiers: Probabilistic Neural Network PNN, Decision Tree (DT), and K-Nearest Neighbor (K-NN). With normal and PVC data from 142 patients, PNN and K-NN achieved the best results. They obtained an average accuracy of 92.25%, sensitivity and specificity of 73.33% and 94.74% respectively. Besides using classifiers, a novel Cardiac Health Index (CHI) was developed by using two nonlinear features: Maximal vertical line (V-max) and Maximal diagonal line (L-max). CHI can be used to indicate whether or not an Electrocardiogram (ECG) signal was taken from patient suffering from PVC. This index enhances the detection of PVC, thereby giving doctors more time to perform other crucial tasks.

• 出版日期2013-9
• 单位天津大学