Effect of segment length on the estimation of magnitude-squared coherence in an experimental model of ventricular tachycardia

The present paper investigates the effect of the segment length on the accuracy of the magnitude-squared coherence as a method able to discriminate monomorphic and polymorphic ventricular tachycardias, when applied to electrograms in a canine model. The magnitude-squared coherence function was estimated using four different segment lengths; 128, 256, 512 and 1024 samples: The highest accuracy (93.1%) was obtained with the segment length of 1024 samples and the lowest (84.3%) with the segment length of 128 samples. Segment length as an element influencing resolution should be taken in consideration for more reliable results.