期刊
CHINESE JOURNAL OF ACADEMIC RADIOLOGY
卷 5, 期 1, 页码 37-46出版社
SPRINGERNATURE
DOI: 10.1007/s42058-022-00087-5
关键词
Clear cell renal cell carcinoma (CCRCC); Computed tomography (CT); International Society of Urological Pathology (ISUP); Radiomics; Machine learning
资金
- National Key Research and Development Program of China [2018YFC1602800]
- Key Project of the National Natural Science Foundation of China [61731009]
Radiomics models can be used to differentiate low and high ISUP grades of CCRCC by analyzing CT image features. The model using NCP and CMP features achieved the highest performance on the independent testing cohort.
BackgroundIdentifying International Society of Urological Pathology (ISUP) grade with noninvasive tools before treatment is important for patients with clear cell renal cell carcinoma (CCRCC).PurposeTo use radiomics models to differentiate low ISUP grade (grades I and II) of clear cell renal cell carcinoma (CCRCC) from high ISUP grade (grades III and IV).Materials and methods156 CCRCC were collected in this study. Shape, first-order, texture, and wavelet-related features were extracted from the non-contrast phase (NCP), corticomedullary phase (CMP), and nephrographic phase (NP) of 3D CT images. We used the intraclass correlation coefficient to select features and logistic regression/supported vector machine to build models. Single-phase models and multi-phase models were built on the training cohort and evaluated them on the independent testing cohort. The receiver-operating characteristic (ROC) curve and area under the curve (AUC) were used for quantification when comparing the performance of each model.ResultsThe model identifying ISUP grades using NCP and CMP features achieved the highest AUC (0.841; 95% CIs = 0.720-0.963) on the independent testing cohort. The AUCs of the multi-phase models were higher than that of the single-phase model.ConclusionRadiomics can be a useful tool in predicting the ISUP score of CCRCC and provide a quantitative diagnosis of CCRCC in clinic.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据