Accurate seed points classification using invariant moments & neural network

Abstract

Segmentation is a key topic in computer vision and medical image processing. Furthermore, it is used in many medical applications and techniques such as registration. Currently, an accurate segmentation is still a challenging task. In this study, the segmentation process starts by selecting seed points within the region of interest. Manual seed points selection can be time consuming and requires an expert to complete the selection. In this paper, we propose a novel method for automatic classification of the seed points in liver Magnetic Resonance Imaging (MRI) belonging to the same patient each time the segmentation is performed. The proposed method uses Geometric Moment Invariants as a feature vector to identify the locations of seed points. Artificial Neural Network (ANN) model is trained using the feature vector of each of the seed points to classify which region of the liver a testing point belongs. We have demonstrated the effectiveness of our technique in classifying three seed points. These seed points represent the left hepatic vein, central hepatic vein, and right hepatic vein of the liver. The proposed method shows high accuracy in classifying the input seed points.