Auxiliary Evaluation Of Bone Age Exploiting TW2 Principle And Convolution Neural Network

Bone age is a quantitative value of the degree of skeletal development.It is a reference for medical diagnosis of adolescent endocrine diseases and prediction of human body height.The bone age is evaluated,according to the characteristics of the developmental stages of the bones.At present,most of bone age evaluations are that doctors manually analyze images of bone and calculate bone age in combination with medical standards of bone age evaluation,this method has the disadvantages of cumbersome,time-consuming and subjective influenced by doctors.A few of bone age evaluations are that computer automatically calculate bone age based on images of bone,this method has the defects that the process and foundation of evaluation are invisible,the result of evaluation can't be changed.In view of the problems in the two ways,thesis explores the method that computer auxiliarily calculate bone age based on the TW2 which is medical standards of bone age evaluation.Thesis studies the bone age evaluation process,hand bone image preprocessing,hand bone segmentation,hand bone feature extraction and hand bone classification which are important techniques in bone age evaluation.Its main research contents and results are as follows:1.Aiming at the problems of invisible process and the unchangeable result of the existing bone age evaluation method,a bone age assisted evaluation framework based on TW2 method is proposed.According to the flow path of bone age evaluation based on TW2,the structure is divided into detecting images of bones,adjusting images of bones,classing the developmental stage of bones,adjusting the developmental stage of bones and calculating bone age.Consequently,the basis and process of the bone age evaluation can be seen by doctors,the result of bone age evaluation can be changed and the accuracy of evaluation is higher.2.Aiming at the tasks of preprocessing image,detecting images of bones and classing the developmental stages of bones which are in bone age evaluation,the method of auxiliary evaluation of bone age based on TW2 method and traditional image processing is studied.The method firstly performs image preprocessing with OTSU threshold and palm connected domain features,and then implements detection of bone based on dynamic threshold and palm anatomy model.Then SVM classifier completes the classifications of bone according to geometrical features and gray distribution characteristics of bone image,finally calculates the bone age based on the mapping rule of bone age in TW2.According to experiments,the method evaluates bone age much faster than doctors and the error of bone age is not much bigger than doctor3.Aiming at the tasks of detecting images of bones and classing the developmental stages of bones which are in bone age evaluation,this paper studies the auxiliary evaluation of bone age based on TW2 method and convolutional neural network.This method firstly builds the detection network of bone image based on the Faster-RCNN network.Then the classification network of the developmental stages of bones is built by tailoring the Inception V1 network.Finally,the bone age is calculated according to all classifications of bones and mapping rules of bone age in TW2.Compared with the method of auxiliary evaluation of bone age based on TW2 method and traditional image processing,the auxiliary evaluation of bone age based on TW2 method and convolutional neural network has lower evaluation error of bone age and faster evaluation speed of bone age by experiments.This method has great potential of application.