Analysis of Wireless Capsule Endoscopy Images using Local Binary Patterns

Abstract

Wireless capsule endoscopy, the gold standard in the screening and diagnosis of small bowel diseases, is one of the most recent investigations for gastrointestinal pathology. This examination has the advantages of being non-invasive, painless, with a large clinical yield, especially for small bowel diseases, but also some disadvantages. The long time necessary for reading and interpreting all frames acquired is one of these disadvantages. This inconvenient could be improved through different methods by using software applications. In this study we have used a software application for texture analysis based on local binary pattern (LBP) operator. This operator detects and removes non-informative frames in a first step, then identifies potential lesions. Our study group consisted of 33 patients from the Gastroenterology and Hepatology Centre Craiova and from the 1st Internal Medicine and Gastroenterology Clinic from the Emergency County Hospital of Craiova. The patients included in the study have corresponded to our inclusion criteria established. The exclusion criteria were represented by the contraindications of the capsule endoscopy. In the first phase of the study, we have removed the non-informative frames from the original videos obtained, and we have acquired an average reduction of 6.96% from the total number of images. In the second phase, using the same LBP operator, we have correctly identified 93.16% of telangiectasia lesions. Our study demonstrated that software applications based on LBP operator can lead to a shorter analysis time, by reducing the overall frames number, and can also provide support in diagnosis.