多层螺旋CT处理技术在脊柱侧弯图像三维重建中的应用

Application of multi-slice spiral CT processing in three-dimensional reconstruction of spinal bend image

现代医学图像,尤其是脊柱侧弯图像的三维重建特征点识别精度较低。对此,在传统三维重建图像基础上应用多层螺旋CT处理技术,进行技术优化。首先通过多层螺旋CT处理技术将三维数据进行投影,建立高斯特征点尺度空间;利用SIFT特征点算法,对尺度空间极值进行检测,从而定位特征点并分配特征点方向,生成关键点描述子,匹配特征向量;为了提高三维图像特征点精度,引入对极几何与基础矩阵,利用多层螺旋CT处理技术进行系统评价,剔除错误匹配点;最后利用多层螺旋CT处理技术的多平面曲面处理技术和相继模型,完成脊柱侧弯图像的相机标定和三维重建。实验数据表明,与传统三维重建技术相比,应用多层螺旋CT处理技术后,三维重建图像稠密点数量明显提高了47%,图像角点匹配率提高25. 5%,明显提高了识别精度。

The recognition accuracy of 3 d reconstruction feature points of modern medical images,especially those of scoliosis images,is relatively low. Based on the application of multi-slice CT processing technology,the 3 d reconstruction technology of PMVS scoliosis images is designed. Firstly,the 3 d data is projected by multi-slice spiral CT processing technology,and the gaussian feature point scale space is established. Through the SIFT feature point algorithm,the extreme value of scale space and the location of feature points are detected,and the direction of feature points is allocated to generate the key point descriptor and match the feature vector. In order to improve the accuracy of feature points of 3 d images,polar geometry and basic matrix are introduced and error matching points are proposed. Finally,the camera calibration and 3 d reconstruction of scoliosis images were completed by using the multi-slice curved surface processing technology and successive model of multi-slice spiral CT. Experimental data shows that compared with traditional 3 d reconstruction technology,PMVS 3 d reconstruction technology has significantly increased the number of dense points in the reconstruction image by 47%,and the image angular point matching rate by 25. 5%,which has significantly improved the recognition accuracy.