VT and VF classification using trajectory analysis

Visual classification of Ventricular Fibrillation (VF) and Ventricular Tachycardia (VT) patterns is a hard task for cardiologists. VT and VF signals are apparently similar in the time domain but their underlying information is totally different. In this paper, an image-based technique is presented which extracts discriminative information from the trajectories of VT and VF signals in the state space. In this way, first, signals are sketched in the state space by the delay time method. Then, the state space is considered as an image and trajectories of VT and VF signals are considered as two different images. The purpose is to design some masks, apply them on the images, and finally classify these masked images by a box counting method. These masks are designed to remove the common information between the two patterns and just discriminative pixels are flagged. After applying the masks, flagged pixels are counted and a threshold is determined through the cross validation phase under the receiver operator curve (ROC) criterion to classify the VT and VF trajectory images. The signals are selected from two different data sets include MIT/BIH and CCU of the Royal Infirmary of Edinburgh. Our experiments show brilliant results which provide 100% classification rate on the training and testing phases. Even, through the cross validation phase, the results remained the same also, so thep value is determined as less than 0.0001, that experimentally shows no over-fitting is occurred.