Bone age assessment based on phalanx radiograms

摘要

To identify the growth status of children, more and more bone age assessment (BAA) systems based on Tanner-Whitehouse (TW) method have been developed to estimate skeletal maturity of children. Therefore, with the concept of TW method, our aim is to develop a useful BAA system for the phalanx radiograms. In this study, the phalangeal bone regions of interest (PROIs) and epiphyseal/metaphyseal regions of interest (EMROIs) are segmented from the two left-hand radiograms. The textural, morphological and shape descriptor features are extracted from EMROIs for the quantization of skeletal maturity. The fuzzy logic rule in conjunction with principal component analysis (PCA) is applied to construct the model of the classification. The database comprising 1550 left-hand radiograms were collected from normal female children aging from 2 to 16 years. For each age, 15 images were used to constitute training data. The classification results by testing on the rest data demonstrate that the proposed BAA system is capable of providing the correct rate of 71.17% with 1 year tolerance and 89.74% with 1.5 year tolerance.