Classification of colonic polyps using Hidden Markov Models

Colonic polyps appear as elliptical protrusions on the inner wall of the colon. Previous algorithms assumed the shape of a polyp to be a spherical cap, so these algorithms are not flexible when the polyps are various cap shapes. This paper proposes an explicit parametric model for the polyps. The model captures the overall shape of the polyp and is used to derive the probability distribution of features relevant to polyp detection. The probability distribution represents the glocal properties of the candidates (where glocal properties capture both the global information and local information of the object). A unit sphere, referred to as a brilliant sphere, is used to represent the glocal information of the polyp. The observation sequence is obtained for the polyp candidates from the brilliant sphere information, and the observation sequence is then assessed by explicit models for classification.