The medical image diagnosis is a difficult task due to small lesions which are not identifiable in an early stage. The early medical diagnosis is important because it increases the mortality rate. This problem is overcome by a Content Base Medical Image Retrieval (CBMIR) system using some medical image databases. A novel method for CBMIR system is proposed in this study. The system classifies the 3-Dimensional (3D) x-ray (hand and lung) image with the two classes. The proposed CBMIR system uses three important stages like (i) pre-processing, (ii) feature extraction and (iii) classification. The normalization technique is used to pre-process the input 3D images at pre-processing stage. The Dual Tree M-Band wavelet Transform (DTMBT) is used for the decomposition of normalized 3D images and also it produces the lower and higher frequency subband coefficients. These subband coefficients are extracted by intensity-based features. These features are stored in the feature database for the classification. The Multiclass-Support Vector Machine (M-SVM) is used for the prediction of different 3D medical database images. The proposed CBMIR system database consists of 3D x-ray images (hand and lung) of two classes for performance evaluation. The proposed CBMIR system yields classification accuracy of 90.95% by using DTMBT based features and M-SVM classifier.
Volume 12 | Issue 3
Pages: 279-286
DOI: 10.5373/JARDCS/V12I3/20201192