Skeletal Bone Age Assessment in Radiographs Based on Convolutional Neural Networks

Skeletal bone age assessment is a common clinical practice to investigate endocrinology, genetic and growth disorders in children. Traditional clinical testing methods are time-consuming and labor-intensive, and there will be operator errors because of the subjective factors of the physicians. The existing automatic bone age detection methods based on automatic extraction of clinical features also has the problems of low accuracy and difficult generalization due to inaccurate feature extraction. In this paper, we propose an end-to-end automatic bone age detection method based on deep learning to process hand bone X-ray images. A Convolutional Block Attention Module (CBAM) is added to the basic model of Inception Resnet v2, the Softmax network layer is changed to the Mean Absolute Error (MAE) index output, and the mean square error loss function is used to evaluate the performance of the bone age detection regression problem. A new stratified k-fold cross validation method is proposed to cover training models on the public dataset of bone age for all races, genders and age ranges. It shows that the MAE between the detected bone age and the labeled bone age is 0.34 years in the results, which is better than the current bone age evaluation method.