光学学报, 2018, 38 (2): 0215005, 网络出版: 2018-08-30
基于机器学习的可降解支架检测与分割算法 下载: 1033次
Detection and Segmentation Algorithm for Bioresorbable Vascular Scaffolds Struts Based on Machine Learning
机器视觉 自动检测与分割 自适应增强算法 可降解支架 血管内光学相干断层扫描图像 machine vision automatic detection and segmentation Adaboost algorithm bioresorbable vascular scaffold intravascular optical coherence tomography image
摘要
针对血管内光学相干断层扫描(IVOCT)成像系统,提出一种改进的自适应增强(Adaboost)算法及一种基于动态规划的轮廓分割算法用于可降解支架的自动检测与分割,实现对支架贴壁情况的自动评估。在检测阶段,利用多层决策树构建Adaboost分类器,实现对支架位置和大小的检测;基于检测结果,利用动态规划算法对支架轮廓进行分割;最后,结合分割结果,对支架贴壁情况进行计算。实验结果显示,所提算法的检测召回率达到91.6%,精确率为87.2%,轮廓分割的平均Dice系数为0.80,表明所提算法能够实现IVOCT影像中可降解支架的准确检测与分割,且具有较好的稳健性。
Abstract
An impoved Adaboost algorithm, together with a profile segmentation method based on dynamic programming (DP), is proposed for automatic detection and segmentation of bioresorbable vascular scaffold (BVS) in intravascular optical coherence tomography (IVOCT) imaging system, to achieve auto estimation on the strut malapposition. During detection, the multi-layer decision tree is applied to the construction of Adaboost classifier, in order to detect the position and size of each strut. Then, the DP algoritm is adopted to segment the struts’ boundaries based on detection results. Finally, combined with the segmentation results, struts malapposition is caculated. Experimental results show that our method reaches the detection recall rate of 91.6% with the precision of 87.3%, and the average Dice coefficient of segmentation is 0.80. It suggests that our method can accurately achieve the detection and the segmentation of BVS struts in IVOCT images, and has high robustness.
鲁逸峰, 金琴花, 荆晶, 陈韵岱, 曹一挥, 李嘉男, 朱锐. 基于机器学习的可降解支架检测与分割算法[J]. 光学学报, 2018, 38(2): 0215005. Yifeng Lu, Qinhua Jin, Jing Jing, Yundai Chen, Yihui Cao, Jianan Li, Rui Zhu. Detection and Segmentation Algorithm for Bioresorbable Vascular Scaffolds Struts Based on Machine Learning[J]. Acta Optica Sinica, 2018, 38(2): 0215005.