Table 3.

Summary and comparison of prior attempts at automated BAA: dataset, method, salient features, and their limitations

Dataset		Method	Features	Limitations
[29]	24 GP female images	SIFT; SVD Fully connected NN	Fixed-sized features vectors from SIFT description with SVD	Training and validation with limited data; deficiency of robustness to actual images
[30]	180 images from [31]	Canny edge detection Fuzzy classification	Morphological features regarding carpal bones	Not applicable for children above 7 years
[32]	205 images from [31]	Canny edge detection Fuzzy classification	Morphological features regarding carpal bones (Capitate Hamate)	Not applicable for children above 5 years for females and 7 years for males
[33]	1559 images from multiple sources	AAM PCA	Features regarding shapes, intensity, texture of RUS bones	Vulnerable to excessive noise in images chronological age used as input
Our work	8325 images at MGH	Deep CNN transfer learning	Data driven, automatically extracted features

SIFT scale invariant feature transform, AAM active appearance model, PCA principle component analysis, SVD singular value decomposition, NN neural network, SVM support vector machine, RUS radius ulna short