3D statistical shape models of radius bone for segmentation in multi resolution MRI data sets

Extracting the structures of interest accurately is one of the main challenges in medical imaging segmentation. Statistical models of shape are a promising approach for robust and automatic segmentation of medical image data. This work describes the construction of a statistical shape model of the Radius bone. For 3-D model-based approaches, however, building the 3-D shape model from a training data set of segmented instances of an object is a major challenge and currently remains an open problem. In this study, we propose an active contour image segmentation method for three-dimensional (3-D) medical images. Our dataset contains T1-weighted images of hand wrist in coronal view. Such images are usually acquired in 9 slices, but we also used 27 slices images in which the spatial resolution is improved by reducing the in depth from 3mm to 1mm. In this study we use 27-slices MRI images to segment radius bone due to their higher resolutions in comparison to 9-slices images. First, using 2D active contour algorithm, radius bone is segmented in coronal slices automatically. Then, a statistical model of radius bone is derived and its mean model is used as the initial mask for 3D active contour algorithm, and 9-slices images are segmented using this algorithm. To compare the 2D and 3D active contour algorithms, 27-slices images are segmented through produced statistical atlas of mean model. Comparison of obtained segmentation and manual segmentation shows that segmentation accuracy in 9-slices images which use mean model will be increased from 75.68% to 91.57%. Acquisition of 9-slicese images takes a shorter time (1/3) in comparison to 27-slices images; therefore, we also derived the statistical model of 9-slices images. In the future works we utilize the proposed approach as part of a computer-aided diagnosis system for bone age estimation.