Abstract: Aim at the problem of traditional surface reconstruction that the efficiency decrease sharply with the promotion of reality, which is also poor of interaction and sensitivity, A new method from CT images for liver lesion surface reconstruction is proposed based on point clouds data. First, the interpolation and adaptive simplification method is improved for CT points cloud; then, two-part reconstruction process is proposed： a coarse base model is firstly created very fast from the distance field through Minimum Energy Constraint （MEC） and the simplified Marching Cube （MC） method, and a linear optimization ARDP （Automatic Robust Directed Projection） method is presented subsequently for automatically calculating the projection vector, which maps the surface points directly on the points cloud. By changing iterative times interactively, the accuracy of model can be improved gradually, so a high-quality model will finally be obtained, which realizes a direct transition from discrete points to the smooth surface. The results show that this way can save the time than that traditional algorithm, with mean error less than 0.000 1, and have good adaptability to irregular lesion models.

Key words: liver lesion, points cloud simplification, MC algorithm, energy constraint, DP algorithm

摘要： 针对传统面绘制方法随真实感的提升效率急剧下降，且交互性及灵敏度较差的问题，基于CT点云数据提出了一种肝脏病灶的表面重建方法。首先改进了点云数据的插值和自适应精简方法；然后提出将模型重构过程分为两部分，先通过最小能量约束和简化的MC算法由点云距离场快速创建粗糙的基底模型，接着提出一种线性最优化的ARDP算法用于自动计算点元投影向量，从而将当前模型表面节点直接映射至点云，通过交互式地确定迭代次数可按需逐步提高模型精确度，最终获取高质量模型，实现散乱点到平滑面的直接过渡。实验结果表明，利用该算法生成平均误差小于0.000 1的高精模型将大大缩短时间，且对不规则病灶模型有着良好的适应性。

关键词: 肝脏病灶, 点云精简, MC算法, 能量约束, DP算法