Aifeng Zhang MS, Presenter: Nothing to Disclose

Arkadiusz Gertych PhD, Abstract Co-Author: Nothing to Disclose

Brent Julius Liu PhD, Abstract Co-Author: Nothing to Disclose

Han K. Huang DSc, Abstract Co-Author: Nothing to Disclose

Sylwia Kurkowska-Pospiech, Abstract Co-Author: Nothing to Disclose

A computer-aided diagnosis (CAD) method has been developed based on features extracted from epiphyseal regions of interest (ROI), which provides accurate bone age assessment of children 12 to18 of age. For children below 12 of age, the features of carpal bone ROI are required to achieve similar degree of accuracy. Past work on carpal bones segmentation has been done using dynamic thresholding. However, due to various stages of carpal bones development and the limitation of the segmentation algorithm itself, feature analysis of carpal bones has not been successful implemented. The goals of this study are: 1) To implement active contour model (snakes) to segment the carpal bones and extract pertinent features, 2) To refine the feature space using the data mining technique, 3) To combine the features from both epiphyseal and carpal ROIs for bone age assessment.

The preprocessing of the hand image was preformed to automatically locate carpal bone ROI. Before an active contour model was applied to segment out carpal bones, prior knowledge about the centers of the bones was needed. The Gibbs random field procedure to locate the center of each carpal bone was developed. The number, size and separation features of all carpal bones were extracted. A feature selection procedure determined the most important features while eliminated the redundant ones. This reduced feature space was used to assess the bone age. The separation of carpal bones was useful for bone age assessment of children 0-9 of age and the amount of bone overlapping for children 9-12.

The new method was tested initially on 30 cases and is being applied to over 500 cases in our collection. Size and shape features of each carpal bone were extracted from each image successfully and applied to bone age assessment.

This research describes an image segmentation method on carpal ROI by active contour model with adaptive parameters. Preliminary results show that the accuracy of bone age assessment of children 0-12 of age is improved with the inclusion of carpal ROI, especially when the epiphyseal ROI analysis originally had failed.

Zhang, A, Gertych, A, Liu, B, Huang, H, Kurkowska-Pospiech, S, Carpal Bone Segmentation and Features Analysis in Bone Age Assessment of Children.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4415569.html