Efficient skeletal bone age estimation method using carpal and radius bone features

Abstract

This study presents design and implementation of an efficient skeletal bone age assessment (BAA) procedure based on features extracted from two dominant wrist bone structures (radius & carpal bones). System works according to renowned Tanner & Whitehouse (TW2) method, based on carpal and epiphyseal region of interest (ROI). It ensures accurate and robust BAA for both girls and boys (age range, 0-10 y). BAA procedure composes of two components, one dealing with carpal bones [Carpal ROI (CROI) sector] and the other with radius bone segments [Radius ROI (RROI) sector]. Input image is initially cropped to extract desired ROI. Anisotropic diffusion filter is employed to reduce noise before extracting features. This is followed by morphological

feature extraction, during which two feature ratios (1 from CROI features and 1 from RROI features) are computed. Feature ratios are used to initially train the system and finally to estimate skeletal bone age. The system was trained with a data set of 120 radiographs (60 males & 60 females) and was tested on a set of 100 radiographs (50 males & 50 females).