Segmentation of structures in medical images: review and a new computational framework

Abstract

Segmentation of structures represented in medical images plays a crucial role in many image processing and analysis applications, like the ones in the medicine and biomechanical areas. In the past few years, the successful implementation of modern mathematical and physical techniques, like level set methods, wavelets analysis, template matching and deformable models, greatly enhances the accuracy of segmentation results. Researchers have proposed many effective algorithms to complete this task.In the first part of our paper, we will make a survey on the state-of-the-art algorithms that can be effectively used for segmenting structures in medical images, explaining their principles, discussing their applications, and enumerate their merits and drawbacks. Due to the different requirement of each possible application and the distinctive characteristics of the structures to be considered, we will focus on the currently most suitable algorithms for the corresponding segmentation task in 2-D and 3-D medical images. Among the referred models, the extensively studied deformable models will be mainly described. We will also discuss the common bottlenecks of these methods when they are used to segment certain structures, like blood vessels, and propose some possible means to handle and overcome these challenges. Our future works in developing segmentation algorithms for medical images, particularly for segment the thin structures of the woman pelvic cavity, will be also addressed.In the second part, we will introduce our computational framework in developing for segmenting structures in medical images. This platform is designed to integrate most of the classical and modern effective algorithms to segment structures in medical images and to be able to integrate the users own algorithms, facilitating their developing, test and fine tune. The features of this platform, such as its interface and the operation procedures available will be described in details. Also, we will illustrate the reasons like why we choose the selected developing tool and the computational libraries integrated and present some experimental results obtained using our platform, namely in structures segmentation and 3D reconstruction. In the end of our paper, we will discuss about our future work under this framework.