Hybrid Technology for Multilevel Feature Extraction in CBMIR

Abstract

A new technique in the healthcare industry called Content-Based Medical Image Retrieval (CBMIR) gives medical personnel an effective way to search for and access pertinent medical images from huge databases. The successful representation and analysis of medical images is made possible by feature extraction techniques, such as the Discrete Wavelet Transform (DWT), Local Binary Pattern (LBP) which are essential to CBMIR systems. As a result, rather than comparing all of the images directly pixel-wise, the retrieval method concentrates on comparing extracted features that capture the most important image attributes. The proposed system is an hybrid approach. A hybrid feature extraction technique combines LBP and DWT and normalizing the features. Next the reduction of the features is done using principal component analysis (PCA). The representation of the images is done using grey-level cooccurrence matrix (GLCM). The study uses a dataset of wrist images to extract the images, next is used to identify the fractures.