A Novel Method for Automatic Detection of Ulcer in Wireless Capsule Endoscopy Images

The Wireless Capsule Endoscopy (WCE) is one among the non-penetrating modern omnipotent detection method mainly used in diagnosing gastrointestinal diseases through direct visualization technique. From the lengthy Wireless Capsule Endoscopy videos of about 8 h duration diagnosing abnormal frames is quiet a difficult task for medical practitioners and we required a time consuming automated detection system for the same. In our proposed automated system using Computer Aided Diagnosis system we are able to detect ulcerous image from the WCE images. Textural feature extraction technique used in this automated system is capable of extracting features not only from one key point but also from nearby of key points. The Gray Level co-occurrence Matrix (GLCM) matrixes are generated for each 16*16 individual patch. In order to improve the classification System accuracy SIFT characteristics method are combined and concatenated with 22 Haralick textural feature. Multilayer perceptron Neural Network can be utilized for perfect classification. Performance of this type of classification method based on features extracted and it accurately identifies ulcerous images with an accuracy of 95.88%.