Analysis of Visual Features in Local Descriptor for Multi-Modality Medical Image

Hizmawati Madzin1, Roziati Zainuddin2, and Nur Sabirin Mohamed3
1Multimedia Department, University Putra Malaysia, Malaysia
2Artificial Intelligence Department, University of Malaya, Malaysia
3Centre for Foundation Studies in Science, University of Malaya, Malaysia

Abstract: In medical application, the usage of multiple medical images generated by computer tomography such as x-ray, magnetic resonance imaging (MRI) and CT-scan images is a standard tool of medical procedure for physicians. The major problems in analyzing various modality of medical image are the inconsistent orientation and position of the body-parts of interest. In this research, local descriptor of texture, shape and color are used to extract features from multi-modality medical image in patches and interest point’s descriptor. The main advantage of using local descriptor is that these features do not need preprocessing method of segmentation and also,  robust to local changes These features are then will be classified based on its modality using support vector machine (SVM) and k-nearest neighborhood (k-NN) classifiers. It shows that different modality have different characteristic and the importance of selecting significance features.

Keywords: Multi-modality medical images, texture, shape and color features analysis, local patches, local interest points.

 

Received November 7, 2011; accepted December 27, 2012

Full Text