Design And Analysis Of Hybrid Classifier For Prediction Of GastroIntestinal Diseases From Endoscopic Images

Abstract

ABSTRACT newlineThe aim of this research work is to predict the five types of gastrointestinal diseases from the 60,136 images of Kvasir Dataset by five phases using the highlighted novel techniques. The first phase starts with noise removal, which is performed using Fuzzy-set based Adaptive Mean Non-Recursive Filter (FAMNRF). An adaptive filtering method by mean is derived for correcting and restoring the corrupted pixel. The second phase is to perform a 3D projection and segmentation. The projection is obtained using Kalman mesh and Semantic segmentation is by the Hue (Colour). The segmentation is done by multiple iterations. During the performance evaluation of the Semantic segmentation enhancer method (SSEM), it is observed that more the number of iterations, higher the segmentation accuracy. The segmentation accuracy increases by 2.58% for every 100 iterations. The research work achieves newline97.3% efficiency for 25,000 iterations. The third phase is for the pattern recognition by using the features of Histogram intensity and Co- occurrence level matrix. The feature selection method used is obtained by merging the Genetic algorithm and the Greedy algorithm as a Hybrid algorithm. The performance evaluation of the Hybrid algorithm for feature selection (HAFS) gives an accuracy of 94.23%. The fourth phase is to identify the abnormal tissue (i.e.) the diseased tissue. An optimal solution classification algorithm (OSCA) is made by using the model newline newline newline newlinematching of the candidate keys. This OSCA provides 95% performance for abnormal tissue identification and a reduced false positive rate of 1.8. The fifth phase is the outcome of all the phases that use machine learning for categorizing the diseases by the three layers Feature classifier layer, Soft-max function, and Disease class layer. There are five classes defined for five diseases like Irritable bowel syndrome, Melanosis coli, Tuber adenoma (Type 1), Ulcerative colitis, Tuber adenoma (Type 2) by the appropriate training set and testing set. This Multiclass classi