基于图像重建的代谢肿瘤总体积分级模型
Total Metabolic Tumor Volume Grading Model Based on Image Reconstruction
宋思良 1陈蔺林 1王泽乾 2吴祎璠 1程紫嫣1
作者信息
- 1. 兰州财经大学,甘肃兰州 730020
- 2. 太原理工大学 现代科技学院,山西 太原 030024
- 折叠
摘要
总体代谢肿瘤体积(TMTV)是一种较为重要的独立于其他指标的预后指标,对患者的准确治疗具有十分重要的指导作用.准确确定TMTV的分级是一项极具挑战的任务,为此文章提出基于图像重建的代谢肿瘤总体积分级模型,其中包含两个模块:分割辅助多维特征学习模块(SAMFL)和重建纠正模块(RCM).前者通过优化和融合分割特征获得更加精确的TMTV;后者采用图像重建和偏差纠正的方法修正分割未能准确识别的区域,从而进一步提高TMTV的准确性.在芝加哥大学医院的数据集上,该模型的准确率达到71%.与其他方法相比,该模型在TMTV分级方面表现得更加出色.
Abstract
Total Metabolic Tumor Volume(TMTV)is an important prognostic indicator independent of other indicators,and it has important guiding role in the accurate treatment on patients.Accurately determining the grading of TMTV is a highly challenging task.Therefore,this paper proposes a total metabolic tumor volume grading model based on image reconstruction,which includes two modules:Segmentation Assisted Multidimensional Feature Learning Module(SAMFL)and Reconstruction Correction Module(RCM).The former obtains more accurate TMTV by optimizing and fusing segmentation features,the latter uses image reconstruction and deviation correction methods to correct the areas that were not accurately recognized in segmentation,thereby further improving the accuracy of TMTV.On the dataset of the University of Chicago Hospital,the accuracy of the model reaches 71%.Compared with other methods,this model performs better in TMTV grading.
关键词
代谢肿瘤总体积/图像重建/图像分割/偏差纠正Key words
total metabolic tumor volume/image reconstruction/image segmentation/deviation correction引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据