Automatic segmentation of carpal area bones with random forest regression voting for estimating skeletal maturity in infants

We evaluate the utility of a model of the structure of the carpal area bones in the hand for predicting skeletal maturity in infants (0 -7 yrs) using a texture based statistical model of appearance. Skeletal maturity assessment is important for diagnosing and monitoring growth disorders. Statistical models of bone shape and appearance have been shown to be useful for estimating skeletal maturity. In this work we locate the carpal bones with an automatic system which uses a Constrained Local Model with Random Forest Regression Voting. Appearance models were built from the segmented images, and texture parameters extracted were used to estimate skeletal age. By analysing the performance on a data-set of 284 digitized radiographs of normal infants we show that an automatically segmented texture based appearance model of the carpal region produces very satisfactory skeletal age estimation results of a mean absolute error of (0.43; 0.53) years, for male and female respectively.