Hybrid Classifier for Bone Age Assessment

Abstract

Bone age assessment (BAA) is a task performed on radiographs by the pediatricians in hospitals to predict the final adult height and to diagnose growth disorders by monitoring skeletal development. Typically, the height and diameter of the wrist were the main indicators considered for judging the bone age. In this research work, it is established that by conducting one-way ANOVA along Procrustes on the shape data of hand X-rays of the persons it is possible to create a discriminate function that would help to classify the geometrical mean shape between a male and a female hand for bone age assessment. The approach explained in this paper also helps in building a better shape model for further extraction of bone parts and at the same time, it helps to build a classifier feature set for supervised hybrid classification for bone age assessment. Primarily this algorithm works on the basis of computing variability of the total shape, between the group’s variability and Procrustes distance. The two-way ANOVA was also conducted to achieve bone shape separations for evaluation purposes. The performance evaluation of the methods was done and its outcome shows a fair degree of accuracy. The work automatically helps to reduce the number of procedures or steps involved in bone age assessment.