Enhanced Hidden Markov Models for accelerating medical volumes segmentation

Abstract

A fully automated unsupervised image segmentation method using Hidden Markov Models (HMMs) is proposed to segment medical volumes. The application of this system to medical volumes has been evaluated using NEMA IE body phantom and a comparison study has been carried out to evaluate HMM and other segmentation techniques which reveal thatHMMdelivers promising results in terms of accurate region of interest detection. Computational time is the main issue to tackle in HMMs, a solution has been proposed and evaluated with respect to the effects of the accelerators on the system accuracy.