Efficient Feature Analysis of Radiographs in Bone Age Assessment

This paper presents the development procedure of feature extraction and analysis of the features extracted from left hand wrist radiographs, which could be further used in estimating the skeletal maturity or bone age. Bone Age Assessment (BAA) is a procedure used in the management and diagnosis of endocrine disorders. It also serves as an indication of the therapeutic effect of treatment. It is of much significance in pediatric medicine in the detection of hormonal growth or even genetic disorders. In the proposed work, the input radiographs are first preprocessed and segmented using a Particle Swarm Optimization (PSO) based segmentation algorithm using Tetrolets, to identify the wrist bones which form the Regions of Interest (ROIs). The identified ROIs are broadly categorized into Carpal bones and Phalangeal bones. From the ROIs, 47 features that describe the morphology of the bones are extracted. Since it will be infeasible for a classifier to handle such a huge number of features, PCA is employed as a tool for feature analysis and thereafter for feature selection. PCA reduces the dimensionality of the features to produce a powerful yet concise feature vector. The results on both the carpal and the phalangeal bones are discussed and the features surpassing the analysis strategy are highlighted.