Segmentation of Humeral head from magnetic resonans shoulder images and determination of Hill-Sachs lesions

Proton density (PD) weighted MR images present inhomogeneity problem, low signal to noise ratio (SNR) and cannot define bone borders clearly. Segmentation of PD weighted images is hampered with these properties of PD weighted images which even limit the visual inspection. The objective of this research is to develop a computer based diagnosis (CBD) system which is capable of recognizing normal and abnormal (edematous and Hill-Sachs deformities) humeral head images by using texture and shape features derived from Hermite transform and PHOG method and determine the effectiveness of segmentation of humeral head from axial PD MR images with active contour without edge (ACWE) model. We extended the use of speckle reducing anisotropic diffusion (SRAD) in PD MR images by estimation of standard deviation of noise (SDN) from ROI. To overcome the problem of initialization of the initial contour of these region based methods the location of the initial contour was automatically determined with use of circular Hough transform. For comparison signed pressure force (SPF) model, Fuzzy C-means and Gaussian mixture models were applied and segmentation results of all four methods were also compared with the manual segmentation results of an expert. The performance of Hermite transform based texture features in classification of humeral bone was compared with curvelet, contourlet and gray level co-occurrence matrix (GLCM) based texture feature descriptors. Hermite based texture feature combined with PHOG (Pyramid of histograms of orientation gradient) which captures the local image shape and its spatial layout. To measure the performance of the extracted features, we deployed MLP (Multi-Layer Perceptron), SVM (Support Vector Machine) and KNN (K- Nearest Neighbors) methods and demonstrated their power in differentiating the normal and abnormal regions. The proposed approach was tested on our own dataset which consists of 79 normal, 140 abnormal (91 edematous and 49 Hill-Sachs) humeral heads in PD weighted MR images. The highest classification accuracy of Hermite based texture analysis and PHOG method was 99.54 % by SVM. Our results suggest that the proposed system is a promising tool for classification of normal and abnormal bone from PD weighted MR images. This study is unique in the literature of using PD weighted MR images and Hermite transform based texture analyses to classify bone lession.