Medical X-ray Images Classification Based on Shape Features and Bayesian Rule
Nooshin J. Fesharaki, Hossein Pourghassem · 2012
The most important stage of search and content retrieval systems of medical images is image classification. The purpose of classification is execution of a process in which a medical image is assigned to a pre-determined class among several classes. In this paper, a classification based on Bayesian rule which makes use of image features in order to classify medical X-ray images is put forward. The main stages of the proposed algorithm are pre-processing, feature extraction, and Bayesian classifier. In the pre-processing stage, in order to reduce the noise and improve the contrast, an adaptive local histogram, a median filter, edge detection filters, thresholding methods, and morphological operators are used for the purpose of clarifying the areas with bones. Subsequently shape features such as Fourier Descriptor, Invariant Moments, and Zernike Moments are extracted from the image. Ultimately, using Bayesian rule, classification is carried out on an X-ray image dataset consisting of 4937 images. The proposed classification algorithm obtains the accuracy rate of 82.87% for a 28-class classification problem.