基于图像识别的腰椎间盘突出症的诊断

Identification and Diagnosis of LDH Based on Medical Image Recognition

为了实现全自动的腰椎MR图像的突出症状分类,提高腰椎间盘突出(LDH)诊断的精确度,提出了一种改进的PSO-SVM分类算法。该方法主要通过使用粒子群算法(PSO)确定SVM的最优参数,提高SVM的分类精度。首先,针对模糊的图像,通过降噪除扰的方法进行预处理。然后,根据椎块和椎间盘的特点,分别使用形状、面积特征和阈值处理进行分割。并采用轮廓极点的方式确定尾椎的四点,提高定位尾椎的精确度。最后,利用改进的PSO-SVM算法对椎间盘突出类型进行分类。通过与传统SVM、WPA-SVM、未改进PSO-SVM算法的对比实验,证明论文改进的算法具有较好的LDH分类效果,验证集、测试集的准确率分别达到92.50%、94.00%。

To realize fully automatic classification of herniation symptoms in lumbar MR images and improve the accuracy of lumbar disc herniation(LDH)diagnosis,an improved PSO-SVM classification algorithm is proposed.Particle swarm algorithm(PSO)is used to determine the optimal parameters of SVM,which improves the classification accuracy of SVM.Firstly,for the blurred image,preprocessing is carried out by the method of denoising and denoising.Then,according to the characteristics of verte-bral mass and intervertebral disc,shape,area features and thresholding are used for segmentation,respectively.The four points of the caudal vertebra are determined by means of contour poles,which improves the accuracy of positioning the caudal vertebra.Final-ly,the type of disc herniation is classified by using the improved PSO-SVM algorithm.Through the comparison experiments with tra-ditional SVM,WPA-SVM and unimproved PSO-SVM algorithms,it is proved that the improved algorithm in this paper has a good LDH classification effect,and the accuracy rates of the validation set and test set are 92.50%and 94.00%respectively.