An Automatic System for Skeletal Bone Age Measurement by Robust Processing of Carpal and Epiphysial/Metaphysial Bones

This paper presents an automatic system for bone age evaluation according to the clinical method of Tanner and Whitehouse (TW2), based on the integration between two systems: The first processes the finger's bones [epiphysis/metaphysis region of interest (EMROI)], whereas the second one processes the wrist bones [carpal region of interest (CROI)]. The system ensures an accurate bone age assessment for age ranges of 0-10 years for males and 0-7 years for females. For both approaches, novel segmentation techniques are proposed. In detail, for CROI analysis, bone extraction is carried out by integrating anatomical knowledge of the hand and trigonometric concepts, whereas a TW2-stage assignment is achieved by combining the gradient vector flow Snakes and the derivative difference of Gaussian filter. One of the main difficulties for bone age assessment based on carpal bones is that trapezium and trapezoid, which are among the biggest bones in the wrist, are often fused even in very young patients. This problem is overtaken by a very effective algorithm that checks the compactness of the identified bones and separates them by using a curvature function. For EMROI analysis, image-processing techniques and geometrical feature analysis, based on the difference of Gaussian, are proposed. The system is evaluated on a set of 106 X-rays, reaching performances of about 90% success rate in bone stage assignment. The system is very reliable and outperforms other effective methods. Moreover, the mean error rate is about 0.46 +/- 0.37 years, which is comparable with clinicians' reliability, for whom the error has been estimated to be 0.33 +/- 0.6 years.