Novel image segmentation algorithms for early detection of osteoporosis from trabecular bone x ray images

Abstract

Osteoporosis is the silent disease which makes the bone brittle and newlineweak leading to mortality. The symptoms of the disease are not noticeable in newlinemany cases until a bone fracture occurs. The non noticeable disease and the newlineseverity and its high mortality rate when encountered are the main reason for newlinemany researches to work on this field. So far researchers have worked on the newlinebony part, cortical bone but to detect the disease at its early stage, the spongy, newlineporous part, the trabecular bone is taken into consideration. Segmentation newlinemethods such as Local Adaptive Thresholding, Watershed Transform, Robust newlineCanny and Laplacian of Gaussian are used for segmenting the newlinemicroarchitecture of trabecular bone.The Separation distance between the newlinetrabecular bones is found using Euclidean distance metrices. newlineThe results are compared with integrated segmentation of Local newlineAdaptive Threshold with Robust Canny, Watershed transform with Robust newlineCanny, Local Adaptive Thresholding with Laplacian of Gaussian and newlineWatershed transform with Laplacian of Gaussian. In Addition to trabecular newlineseparation distance, Execution times of the algorithms are analyzed. The newlineIntegration of Watershed transform with Laplacian of Gaussian yielded a newlinebetter result over the other methods. This was capable of detecting smallest newlinepores when compared with other single segmentation method and integration newlinemethods. Moreover from the result obtained it is seen that Gabor wavelet newlinetransform used in image enhancement process yields a better trabecular newlineseparation distance values .However the processing time is slightly greater newlinewhich can be neglected when comparing the efficiency of detection. newlineIntegration of Region-based and Boundary-based Segmentation techniques newlinecan further be combined to process real time analysis of the disease. newline