Seeds classification for image segmentation based on 3-D affine moment invariants

Abstract

Purpose

Image segmentation is a crucial topic in computer vision and medical image processing. However, accurate image segmentation is still a challenging task for many medical applications. The region growing based image segmentation process starts by selecting seed points within the region of interest. Hence, the segmentation algorithm is sensitive to the initial seeds and the result can be influenced greatly by the accuracy of seed selection process. Manual seed selection can be time-consuming and requires an expert to complete the selection. In this paper, we propose an innovative approach to automating the initialization process of the liver segmentation of magnetic resonance images. The seed points, which are needed to initialize the segmentation process we proposed in [1], are extracted and classified by using affine invariant moments and artificial neural network.

Methods

We calculated eleven invariant moments for 56 different points within the region of interest of an abdominal MR image. These points represent the bifurcation points of the vessels centerlines of the liver. In this paper, we divide the shape of the liver into four regions; left hepatic vein, center hepatic vein, hepatic portal vein, and right hepatic vein. Then, the moments are classified by an artificial neural network to decide to which part of the liver each point belongs.

Results

We have validated our proposed technique by comparing the method with manual seed selection. The experimental results show that our method outperforms the manual method in terms of the accuracy of seed point selection and the speed of the process.

Conclusions

The proposed technique is considered a robust technique for 3D point selection and classification. The selected seed points are used to initialize the segmentation process. The aim of this method is to efficiently detect and identify the seed points in MR images.

http://ift.tt/2hCs62s