Automatic Image Segmentation And Cobb Measurement Of Spine Base On U-Net

Adolescent Idiopathic Scoliosis is a common spinal surgery disease.Because the cause of the disease is unknown,it cannot be targeted for treatment according to the cause.Therefore,early detection and intervention are important particularly.The Cobb angle of scoliosis is an important reference for measuring the condition of scoliosis.At present,the measurement of Cobb angle needs to be performed by a spine surgeon manually,and the measurement process is slowly and prone to errors.Therefore,develop an algorithm to measure the Cobb angle of spine images quickly and fully automatically it is of great practical significance.Cobb is the angle of scoliosis and one of the reference standards for the severity of scoliosis.At present,measurement of Cobb angle in clinical,professional doctors draw two reference lines on the image according to their experience,and then use a protractor to measure the angle,which is inefficient and prone to errors.Because different doctors have different diagnostic experience,measuring the same image may produce different results.In order to solve the above problems,this paper proposes a fully automatic measurement method of scoliosis Cobb angle based on the U-Net model.Segmenting the spine X-ray images based on the deep learning model,and then identify contour and determination boundary of the segmentation result.Finally,calculate the Cobb angle based on the upper and lower boundaries The main research contents of this paper are as follows:(1)Segmenting spine X-ray images based on deep learning models.Collect standing spine anteroposterior X-ray images,filter the data to remove unsharp images,then preprocess and augment the remaining data,make model training labels finally.Build a convolutional neural network model based on the U-Net model structure,make improvements to the model,improve the segmentation accuracy and algorithm speed,train the network and save the model parameters finally.(2)Recognition the spine contour on the segmentation result,and design an algorithm to calculate Cobb from contours automatically.Perform topological analysis on the segmented binarized image,determine the surrounding relationship of the boundary,calculate the contour information of each vertebra,and then calculate the minimum circumscribed rectangle base on the contour.Determine the upper and lower boundaries of the vertebra by the four corners of the rectangle,and calculate the maximum Cobb angle according to the upper and lower boundaries finally.(3)Integrate all algorithmic processes and program a visual software operation interface.Program a software interface that can be used quickly and easily.Users without programming experience can use the software to measure the Cobb angle quickly.The software can clearly display the measurement results and has functions such as saving the measurement results.