Content Based Medical Image Retrieval for Histopathological, CT and MRI Images
A content based approach is followed for medical images. The purpose of this study is to access the stability of these methods for medical image retrieval. The methods used in color based retrieval for histopathological images are color co-occurrence matrix (CCM) and histogram with meta features. For texture based retrieval GLCM (gray level co-occurrence matrix) and local binary pattern (LBP) were used. For shape based retrieval canny edge detection and otsu‘s method with multivariable threshold were used. Texture and shape based retrieval were implemented using MRI (magnetic resonance images). The most remarkable characteristics of the article are its content based approach for each medical imaging modality. Our efforts were focused on the initial visual search. From our experiment, histogram with meta features in color based retrieval for histopathological images shows a precision of 60 % and recall of 30 %. Whereas GLCM in texture based retrieval for MRI images shows a precision of 70 % and recall of 20 %. Shape based retrieval for MRI images shows a precision of 50% and recall of 25 %. The retrieval results shows that this simple approach is successful.
Content based image retrieval, medical imaging, medical imaging modalities, color, texture, shape