利用群体CT计划图像的多任务前列腺自动分割

发布时间：2018-04-20 05:23

  本文选题：前列腺分割 + CT计划图像　； 参考：《应用科学学报》2017年01期

【摘要】：为了实现CT计划图像中前列腺的自动分割,提出一种基于群体CT计划图像的多任务前列腺分割方法.将群体CT计划图像分别映射到不同参考图像空间,形成多个训练任务.利用随机森林算法和自动上下文模型训练出一系列随机森林分类器,将分类器作用在待分割CT计划图像上获得多个分类概率图,最后使用多数投票法求得最终分割结果.实验表明,与单任务分割方法相比,基于群体CT图像的多任务分割能有效提高CT计划图像中前列腺的分割准确率.
[Abstract]:In order to realize automatic prostate segmentation in CT planning image, a multitask prostate segmentation method based on colony CT planning image is proposed. The group CT planning images are mapped to different reference image spaces to form multiple training tasks. A series of random forest classifiers are trained by using the stochastic forest algorithm and the automatic context model. The classifier is used to obtain multiple classification probability maps on the CT planned images to be segmented. Finally, the final segmentation results are obtained by majority voting method. Experimental results show that multitask segmentation based on colony CT image can effectively improve the accuracy of prostate segmentation in planned CT images compared with single-task segmentation method.
【作者单位】： 南京邮电大学地理与生物信息学院;南京邮电大学通信与信息工程学院;
【基金】：国家自然科学基金(No.31671006)资助
【分类号】：R737.25;TP391.41
，

本文编号：1776436