目的 对虚拟内窥镜心脏超声图像滤波和图像分割方法进行评价。方法 对30例小儿心脏三维超声数据进行滤波和图像分割，分别采用光线投射法、光线投射法（梯度阈值3～19）、光线投射法（梯度阈值3～6）及滤波并分类后的光线投射法（梯度阈值3～6），对显示结果进行比较。结果 显示效果较好者为滤波并采用聚类法分类后光线投射法，其显示的结果（如二尖瓣、三尖瓣以及房间隔缺损）十分清晰。结论 对虚拟内窥镜心脏超声图像预处理时，宜采用滤波、聚类法分类后光线投射法。

Objective To assess filtering and segmentation of three-dimensional echocardiographic images by virtual endoscopy. Methods Filtering and segmentation were performed for three-dimensional echocardiographic images of 30 children. Three-dimensional echocardiographic heart models were reconstructed by four different methods: ray-casting algorithm, ray-casting algorithm （gradient threshold ranged 3-19), ray-casting algorithm （gradient threshold ranged 3-6）and ray-casting algorithm (visualization after filtering and classification, gradient threshold ranged 3-6). Visualization results were compared. Results Ray-casting algorithm with visualization after filtering and classification performed better than the other methods, and the visible heart models showed precisely anatomy structure and spatial relationship including mitral valve, tricuspid valve and atrial septal defect. Conclusion Filtering and classification is helpful for visualization of three-dimensional echocardiographic virtual endoscopy.