Research And Implementation Of Bone Age Recognition System Based On Deep Learning

Bone age detection plays an important role in medical health,sports,judicial expertise and other fields.The traditional bone age recognition and detection is based on the doctor’s manual interpretation of hand bone X-ray image.This method is subjective and needs experience,and there are some errors.Computer aided detection can effectively improve the efficiency of medical diagnosis.Using deep learning method to classify and evaluate bone maturity can effectively improve the efficiency and accuracy of detection results.In this paper,the process of hand bone X-ray image location and preprocessing based on feature extraction and enhancement is proposed,and the bone age recognition and detection algorithm based on deep learning and convolution neural network is proposed,and the corresponding automatic recognition system is realized.The main research contents and achievements of this paper include:(1)Localization and preprocessing of X-ray image of hand bone.According to the characteristics of RSNA2017 experimental data set,a set of preprocessing process of hand bone X-ray image is designed.Histogram feature enhancement technology is used to improve the image contrast;Canny algorithm is used to extract the hand shape from the hand bone image,and U-NET network is used to train and extract the hand shape from the large data set.The test on the above data set shows that the preprocessing process of hand bone X-ray image is effective.(2)In this paper,the depth learning method is used to detect the bone age of X-ray images of hand bone.Resnet is introduced,i.e.,the first part of Inception Resnet V2 network is used to replace the convolutional layer in Faster R-CNN,and the improved Faster R-CNN is used to detect bone age.The optimal model is obtained through training.The average error of bone age detection is 0.583 years when the learning rate is 0.001,which is better than the widely used methods and the new methods proposed by experts and scholars at home and abroad.At the same time,the control experiment of gender quantity set was set in the experiment to detect the influence of gender on the detection results of bone age recognition.Using the above model,an automatic bone age recognition and detection system is designed and implemented,which is deployed on the server,so that users can easily predict the bone age of hand bone X-ray images.The test shows that it meets the requirements.