Stomach Disorder Detection and Analysis using Hybrid Learning Vector Quantization with African Buffalo Optimization Algorithm

Abstract

The human digestive system's electrical activity may be recorded noninvasively by Electrogastrography (EGG). Electrogastrograms are recordings of the electrical activity produced by the stomach muscles. EGG Several gastrointestinal disorders may be diagnosed and their severity measured using EGG signal properties. The literature has several contributions to the categorization of EGG signals. The majority of them make use of either the EGG's frequency or time data. The wide variety of EGG signals is a challenge for current automated categorization methods. Therefore, this study's objective is to develop a lightweight classifier that achieves high classification accuracy while using little processing resources. To acquire normal and abnormal EGG signals at a reasonable cost, a three-electrode measuring device is created here, with classification performed by a hybrid of Linear Vector Quantization and the African Buffalo Search Algorithm (HLVQ-ASO). The results show that the information richness of recorded EGG signals from healthy persons is greater for EGG signals captured using a surface electrode with a contact diameter of 19 mm as compared to 16 mm. To demonstrate their validity and degree of classification accuracy, the results computed using the suggested classifiers are compared with the current classifiers like Artificial Neural Network, Multimodal Support Vector Machine (MSVM), and Improved Convolutional Neural Network (CNN). Additionally, the HLVQ-ASO-based classification method is effective in differentiating between normal and diabetic EGG signals, found a sensitivity of 97% and a specificity of 98.8%. For a dataset of 500 samples, the classification accuracy is 97%.