Bone Age Assessment Using Support Vector Machine Regression

Bone age assessment on hand radiographs is a costly and time consuming task in radiology. Recently, an automatic approach combining content-based image retrieval and support vector machines (SVM) has been developed. In this paper, we apply support vector regression (SVR) as a novel method, yielding a gain in performance. Our methods are designed to cope with the age range 0–18 years as compared to the age range 2–17 of the commercial product BoneXpert. On a standard data set from University of South Carolina, our approaches reach a rootmean-square error of 0.95 and 0.80 years for SVM and SVR, respectively. This is slightly below the performance of the commercial product using an active shape approach.