基于支持向量机与随机游走结合的GGO型肺结节分割方法

目的:为了提高磨玻璃型肺结节（GGO型肺结节）的分割精度，提出一种基于支持向量机与随机游走相结合的分割方法。方法:利用已手动分割的GGO型肺结节训练支持向量机。由训练后的分类模型在待分割的GGO型肺结节图像中选择种子点，然后利用随机游走算法根据支持向量机选取的种子点进行GGO型肺结节图像分割。结果:该研究纳入150个待分割GGO型结节图像，上述分割算法的平均准确率为98.05%、平均召回率为96.35%和平均F1值为98.05%。与传统方法相比，本方法实现了GGO型肺结节的精确自动化分割，对GGO型肺结节定量、定性分析提供更加准确的客观依据。结论:该方法利用支持向量机选取种子点，并利用随机游走进行结节分割可以有效地对GGO型肺结节进行分割，具有简单高效，准确率高的优点。

GGO pulmonary nodule segmentation method based on support vector machine and random walk

Objective: To improve the segmentation precision of the ground glass opacity(GGO) type pulmonary nodules, by a segmentation method based on the combination of support vector machine and random walk. Methods: GGO pulmonary nodules after manual segmentation were used to train support vector machines. The trained classification model was used to select seed points on the image of GGO pulmonary nodules that need to be segmented. The random walk algorithm segmented was applied to segment the GGO pulmonary nodules based on the seed points selected by the support vector machine. Results: The study included 150 GGO pulmonary nodules to be segmented. The average accuracy rate of the above algorithm was 98.05%, the average recall rate was 96.35% and the average F1 value was 98.05%. Compared with the traditional method, the algorithm could achieve accurate segmentation of GGO pulmonary nodules and provide more accurate and objective basis for the quantitative and qualitative analysis of GGO pulmonary nodules. Conclusion:The method relies on support vector machines to select seed points and random walk to segment . This method can effectively segment GGO lung nodules, and has the advantages of simplicity, high efficiency and high accuracy.