基于光流场模型的弹性配准在放射治疗中的应用

目的 采用Matlab软件编程,实现基于光流场模型的图像弹性配准,并将其应用于放射治疗患者CT图像中,为进一步分析肿瘤及危机器官的变化提供软件支持.方法 对接受放射治疗的宫颈癌患者的盆腔CT图像,分别基于Lucas-Kanade与Horn-Schunck光流场模型实施弹性配准.结果 Lucas-Kanade模型弹性配准前后对比,最小均方误差(MSE)减少59.80％(19.5915/32.7590),相关系数(CC)提高6.17％(0.0331/0.9369);Horn-Schunck模型弹性配准前后对比,最小MSE减少67.75％(22.1940/32.7590),CC提高6.36％(0.0596/0.9369).结论 上述两种基于光流场模型的弹性配准取得较好的配准效果,且耗时较少；但其精度对于细微差别仍显不足,需进一步提高形变的精确性.

Elastic registration based on optical flow model and its application in clinical radiotherapy

Objective To achieve elastic registration based on optical flow model using Matlab software program, and apply it in radiotherapy CT image, in order to provide software support for further analysis of the tumor and the crisis in organ changes. Methods Based on the Lucas-Kanade and Horn-Schunck optical flow model, elastic registration was implemented on pelvic CT images of patient with cervical cancer. Results Compared with images before elastic registration, mean square error (MSE) decreased by 59.80% (19.5915/32.7590) after elastic registration based on the Lucas-Kanade model, while correlation coefficient (CC) increased by 6.17% (0.0331/0.9369). For Horn-Schunck model, MSE decreased by 67.75% (22.1940/32.7590), and CC increased by 6.36% (0.0596/0.9369). Conclusion Both Lucas-Kanade and Horn-Schunck optical flow model can be used to obtain elastic registration costing less time, but lack of precision for appearing small differences, and their accuracy need to be further improved.