Bone Age Assessment with X-Ray Images Based on Contourlet Motivated Deep Convolutional Networks

Xun Chen,Chao Zhang,Yu Liu

DOI: https://doi.org/10.1109/mmsp.2018.8547082

2018-01-01

Abstract:Bone age assessment (BAA) is a widely performed procedure for skeletal maturity evaluation in pediatric radiology. It has various clinical applications such as diagnosis of endocrine disorders, monitoring of growth hormone therapy and prediction of final adult height for adolescents. Recent studies indicate that deep learning techniques have great potential in developing automated BAA methods with significant improvements in terms of conventional computer-assisted approaches. In this paper, we propose a multi-scale feature fusion framework for bone age assessment based on deep convolutional neural networks. In our method, the non-subsampled contourlet transform (NSCT) is firstly performed on an input left-hand radiograph to obtain its multi-scale and multi-direction representations. Then, the decomposed bands at each scale are fed to a convolutional network that contains a series of convolutional and pooling layers for feature extraction, respectively. Finally, the feature maps from different branches are concatenated and put into a regression network consisting of several fully connected layers to obtain the bone age estimation. Experimental results on a public BAA dataset demonstrate that the proposed method can achieve state-of-the-art performance.