Classification of hand bones for bone age assessment

This paper presents a new approach to classifying bones of the hand-wrist images into pediatric stages of maturity using point distribution models (PDM). The method consists of two phases: the training phase and the classification phase. During training, examples of bones from each class are collected so that the allowable shape deformations for each class are learnt. A model representing each class is generated. These models are subsequently used to classify new examples of the bones. During classification all models are compared to the input image and the object is assigned to the class whose model is the closest match. Experimental results obtained using 120 images of the third distal and middle phalanxes showed the usefulness of the method for classifying these bones into their proper stages of maturity.