Classification of Gas Bubble in A Doppler Ultrasound Signal Using Synchrosqueezing Transform

Abstract

This paper presents classification of gas bubble in a Doppler ultrasound signal using Synchrosqueezing Transform (SST). The SST decomposes the signal into a number of scales. In this research work, initially two statistical parameters, the peak value and variance are estimated to Figure out the scales that contains gas bubbles. Then the signal is reconstructed from the coefficient values within the selected scale. Some parameters are defined and calculated from the reconstructed signal. These parameters are used to classify gas bubble signal using naïve Bayes classifier. However, two classes “bubble” and “not bubble” are identified by training data sets. Therefore, on the basis of posterior probability, the class of the signal is defined. Finally, performance of gas bubble detection is evaluated in terms of sensitivity and positive predictivity tests. Our proposed method is applied on grade 0, I, II, and III signals. It is observed that, good classification result is achieved in grade I and grade II. In grade 0, no gas bubble is found. In the experiment, 92% gas bubble is classified in grade I, 84% gas bubble is classified in grade II and 80% gas bubble is classified in grade III. Experimental result shows that the proposed method achieves better accuracy than the conventional method in the literature.