Bone age evaluation method based on non-subsampled contour wave transform and convolution neural network

Abstract

The invention discloses an X-ray image bone age evaluation method based on non-downsampled contour wave transform and convolution neural network, includes such steps as non-downsampled contour wave transformation of normalized size X-ray image, The high-frequency directional subbands and a low-frequency coefficient map at several scales are obtained, and then they are input into a multi-channel convolutional neural network to obtain the characteristic maps at different scales. Finally, these characteristic maps are stacked together and input into a regression network composed of several connected layers to obtain the bone age prediction values. The above process is implemented in an end-to-end network structure, and error back propagation mechanism is used to achieve network training. Themethod of the invention utilizes the non-down sampling contour wave transform to pre-extract and separate the features of the original spatial domain image, can overcome the difficulty of network training of the existing depth learning method on the small-scale data set, improve the generalization performance of the network, and thus provides a more accurate and reliable evaluation method for clinical application.

Description

Technical field [0001] The invention relates to the field of medical image analysis, in particular to a bone age assessment method based on non-down-sampling contourlet transform and convolutional neural network. Background technique [0002] As a commonly used technique in pediatric radiology, bone age assessment is mainly used to judge the maturity of children's skeletal development. It is of great significance to the diagnosis of endocrine disorders, the monitoring of growth hormone therapy, and the prediction of the final height of adolescents. At present, the most popular clinical bone age assessment method is based on the X-ray images of the subject's left hand and wrist. The doctor completes the bone age assessment of the subject by analyzing the development of the corresponding bone structure in the X-ray image. However, the accuracy of bone age assessment results depends heavily on the doctor's experience and level. The assessment results of different doctors may vary gr...

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the problems existing in the prior art, the present invention provides a bone age assessment method based on non-subsampling contourlet transform and convolutional neural network, in order to solve the problem of network analysis on small-scale data sets in existing methods based on deep learning. Training difficult problems, improving the generalization performance of the network, thus providing a more accurate and reliable evaluation method for clinical applications

Images