1. Novel radiomics signature predicts lymph node metastasis in T1 colorectal carcinoma
- Author
-
Tianye Niu, Yanlei Ma, Sheng Zhang, and Xinxiang Li
- Subjects
Oncology ,Cancer Research ,medicine.medical_specialty ,business.industry ,Colorectal cancer ,Lymph node metastasis ,medicine.disease ,Metastasis ,Radiomics ,Internal medicine ,medicine ,Lymph ,business - Abstract
506 Background: This study evaluates the predictive performance of radiomic features in metastasis of T1 colorectal carcinoma (CRC) to lymph nodes. Methods: A total of 10 200 CRC patients from our clinical cancer center included in this analysis. 225 eligible cases diagnosed with T1 CRC were included and divided into two groups: computed tomography (CT) image group (n = 82) and magnetic resonance image (MRI) group (n = 143) based on the preoperative image data available. A total of 548 radiomic features were extracted from each case and analyzed, and then a panel of radiomic features associated with lymph node metastases (LNM) were selected using Mann-Whitney U test. Combining these selected radiomic features and clinical data, the predictive performance for LNM was calculated using receiver operating characteristic (ROC) curves. Results: The prediction accuracy for LNM of T1 CRC could be improved to 0.88 by area under the receiver operating characteristic curve (AUC) through integration of one radiomic feature and three clinical indicators in CT group. In the group of contrast enhanced T1-weighted MRI (T1w-MRI), combination of two radiomic features and three clinical parameters present an AUC value of 0.85. In the group of T2-weighted MRI (T2w-MRI), combination of four radiomic features and five clinical characteristics identified T1 tumors with LNM with an AUC value of 0.87. Conclusions: The current study present a good predictive performance of combination of radiomic features with clinic characteristic in identifying T1 CRC with LNM, which may provide an important opportunity for us to make clinical treatment decision-making for T1 CRC patients.
- Published
- 2019