A study on the radiomic correlation between CBCT and pCT scans based on modified 3D-RUnet image segmentation

Yanjuan Yu, Guanglu Gao, Xiang Gao, Zongkai Zhang,Yipeng He,Liwan Shi,Zheng Kang

FRONTIERS IN ONCOLOGY(2024)

引用 0|浏览2
暂无评分
摘要
Purpose The present study is based on evidence indicating a potential correlation between cone-beam CT (CBCT) measurements of tumor size, shape, and the stage of locally advanced rectal cancer. To further investigate this relationship, the study quantitatively assesses the correlation between positioning CT (pCT) and CBCT in the radiomics features of these cancers, and examines their potential for substitution.Methods In this study, 103 patients diagnosed with locally advanced rectal cancer and undergoing neoadjuvant chemoradiotherapy were selected as participants. Their CBCT and pCT images were used to divide the participants into two groups: a training set and a validation set, with a 7:3 ratio. An improved conventional 3D-RUNet (CLA-UNet) deep learning model was trained on the training set data and then applied to the validation set. The DSC, HD95 and ASSD were calculated for quantitative evaluation purposes. Then, radiomics features were extracted from 30 patients of the test set.Results The experiments demonstrate that, the modified model achieves an average DSC score 0.792 for pCT and 0.672 for CBCT scans. 1037 features were extracted from each patient's CBCT and pCT images, 73 image features were found to have R values greater than 0.9, including three features related to the staging and prognosis of rectal cancer.Conclusion In this study, we proposed an automatic, fast, and consistent method for rectal cancer GTV segmentation for pCT and CBCT scans. The findings of radiomic results indicate that CBCT images have significant research value in the field of radiomics.
更多
查看译文
关键词
radiomics,cone-beam CT,rectal cancer,3D-RUnet,CLAHE
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要