Research On Bone Age Recognition Method Based On CHN

Bone age is an important index for youngsters’ growth measurement, which is widely used in disease diagnosis, sports selection and administration of justice and other areas because of its stability and reliability. The wrist is able to reflect the growth and maturation level of human body skeleton comprehensively, meanwhile it is easy to get the wrist X-ray image that get less radiation, therefore the wrist is recognized the optimal skeleton for bone age recognition. Traditional bone age recognition evaluated artificially has strong subjectivity, long time measurement and high professional level. With the development of computer technology, putting digital image processing technology into the field of bone age to realize automatic bone age recognition is the hot spot in bone age researches at present. This paper has studied a method of bone age recognition based on CHN. According to the characteristic of the wrist X-ray image, the paper carried out research on the automatic bone age recognition process. Detail research contents and experiment results were as follow:(1) Proposed a phalanges segmentation algorithm based on k-cosine curvature. First, carry operation on the wrist X-ray image with star median filter and image binarization. Then carry on the connective regional mark and selection, dilation, erosion et al. morphology operations to the binary image, in order to obtain a complete binary image of wrist. In the end, the key points of phalanges are located by algorithm of k-cosine curvature to segment the fingers on the wrist X-ray image efficiently.(2) With the character of the wrist X-ray image, epiphysis/metaphysis region of interest (EMROI) are extracted based on gray level profile on the central axis of each finger. Integrate algorithm of difference of Gaussian filtering and Canny operator edge extraction, which are applied to the EMROI processing, and carry on the binarization and hole filling et al. operations to it. Then the essential features which can characterize each grade of epiphyses during growth and development are extracted and analyzed, and take it as the feature parameters of bone age for the next bone age recognition.(3) Proposed a bone age recognition algorithm based on wavelet support vector machine (WSVM). In combination with classification algorithm of pattern recognition, a variety of classification methods which used to recognize image are compared. Then the fundamentals and classification algorithm of support vector machine (SVM) are analyzed. Finally, based on wavelet kernel function, the bones are classified into different grades using algorithm of WSVM muti-classification, and are compared with the tradition algorithm of SVM muti-classification. The experiments show that WSVM have registered a remarkable elevation in both nonlinear approximation and recognition accuracy, verify the effectiveness and feasibility of this algorithm.