Detecting the Functional Gastrointestinal Disorder based on Wavelet Tra

In recent years, researchers have developed powerful wavelet techniques for the multi-scale representation and analysis of signals [1][2][3][4][5]. Wavelets localize the information in the time-frequency plane[6]. One of the areas where these properties have been applied is diagnosis. Due to the wide variety of signals and problems encountered in biomedical engineering, there are various applications of wavelet transform [7][8][9][10]. Like in the heart, there exists a rhythmic electrical oscillation in the stomach. With the accomplishment of the whole digestive process of the stomach, from mixing, stirring, and agitating, to propelling and emptying, a spatiotemporal pattern is formed [11]. The stomach has a complex physiology, where physical, biological and psychological parameters take part in, becoming difficult to understand its behavior and function. It is presented the initial concepts of a mechanical prototype of the stomach, it uses to describe mechanical functions of storing, mixing and food emptying [12][13]. The nature of gastric electrical activity in health and disease is fairly well understood. In man, it consists of recurrent regular depolarization (slow waves or electrical control activity-ECA) at 2.5 to 4 cycles per minute, and intermittent high-frequency oscillations (spikes or electrical response activity-ERA) that appear only in association with contractions. The oscillations commence at a pacemaker site high up in the corpus and propagate to terminate at the distal antrum. The velocity of propagation and the signal amplitude increase as the pylorus is approached. ECA are the ultimate determinant of the frequency and direction of propagation of phasic contractions, which are responsible for mixing and transp... ...ls from their wavelet coefficients, before they are applied to a static neural network for further classification. The design of neural network is simple because only interesting features of GEA types are presented. The experimental results show that it’s possible to classify GEA types by using this simple neural network architecture. We present the results from a network which is trained on sample types. The approach of classifying the output of a feature detector offers greater computational efficiency and accuracy than that of attempting to use a neural network directly upon a GEA signal, and yet preserves the ability to train and flexibility of a neural network. Section 3 of this paper describes the architecture of a network to classify the GEA types for detecting abnormalities. Experimental results of training and testing a network are presented in section 6.