A novel endoscopic image analysis approach using deformable region model to aid in clinical diagnosis

A novel endoscopic image analysis approach using deformable region model has been proposed for clinical diagnosis. The deformable region model approach determines the lumen from the endoscopic image by giving an approximate boundary plan of the lumen by using minimum cross-entropy algorithm, and then deforming this plan to the real boundary automatically. A growing and shrinking method is employed for deforming the boundary. Connected components labeling method is used to eliminate small regions outside the lumen. The algorithm uses the grassfire concept implemented recursively. Applying connected components labeling algorithm on the inverted image eliminates small regions inside the lumen. Subsequently, deriving relevant features from the lumen information, abnormalities in the endoscopic image can be detected. The proposed scheme has been tested using a number of endoscopic images. The results obtained support the feasibility of the proposed approach and are of great interest and potential diagnostic value to the clinicians.