Background and Objective: Capsule Endoscopy (CE) is a non-invasive, patient-friendly
alternative to conventional endoscopy procedure. However, CE produces 6 to 8 hrs long video posing a tedious challenge to a gastroenterologist for abnormality detection. Major challenges to an
expert are lengthy videos, need of constant concentration and subjectivity of the abnormality. To
address these challenges along with high diagnostic accuracy, design and development of automated
abnormality detection system is a must. Machine learning and computer vision techniques are devised to develop such automated systems.
Methods: Study presents a review of quality research papers published in IEEE, Scopus, and Science
Direct database with search criteria as capsule endoscopy, engineering, and journal papers. The initial search retrieved 144 publications. After evaluating all articles, 62 publications pertaining to image analysis are selected.
Results: This paper presents a rigorous review comprising all the aspects of medical image analysis
concerning capsule endoscopy namely video summarization and redundant image elimination, Image
enhancement and interpretation, segmentation and region identification, Computer-aided abnormality detection in capsule endoscopy, Image and video compression. The study provides a comparative
analysis of various approaches, experimental setup, performance, strengths, and limitations of the
aspects stated above.
Conclusions: The analyzed image analysis techniques for capsule endoscopy have not yet overcome
all current challenges mainly due to lack of dataset and complex nature of the gastrointestinal tract.