Value of multi-sequence MRI radiomics combined with KRAS mutation nomogram model in predicting the sensitivity of neoadjuvant chemotherapy in patients with rectal cancer
10.3760/cma.j.cn112149-20240327-00162
- VernacularTitle:多序列MRI影像组学联合KRAS突变列线图模型预测直肠癌患者新辅助放化疗敏感性的价值
- Author:
Hongbo HU
1
;
Ying ZHANG
;
Sheng ZHAO
;
Hao JIANG
;
Xue LIN
;
Huijie JIANG
Author Information
1. 哈尔滨医科大学附属第二医院影像科，哈尔滨 150086
- Keywords:
Rectal Neoplasms;
Magnetic resonance imaging;
Radiomics;
KRAS gene
- From:
Chinese Journal of Radiology
2024;58(10):1069-1074
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct a multi-sequence MRI radiomics combined with KRAS mutation nomogram model to predict the efficacy of pathological complete response (pCR) in patients with rectal cancer after neoadjuvant chemoradiotherapy.Methods:This study was a case-control study. A total of 126 patients with rectal cancer who were treated with neoadjuvant chemoradiotherapy in the Second Affiliated Hospital of Harbin Medical University from June 2020 to December 2023 were retrospectively collected. The pathological response of the postoperative specimens was graded, with 64 cases of pCR and 62 cases of non-pCR. KRAS gene detection was performed on the pathological tissues before neoadjuvant chemoradiotherapy. Among the patients, 34 cases had KRAS mutants and 92 cases had KRAS wild-types. The 126 patients were randomly divided into a training set and a validation set at a ratio of 8∶2 by the random number method, with 101 and 25 cases, respectively. The difference in KRAS mutation status between the pCR group and the non-pCR group was compared by the χ2 test. The radiomic features were extracted from the baseline T 2WI, diffusion-weighted imaging, and apparent diffusion coefficient images of the patients. The optimal radiomic features were screened out to establish the radiomics model. The radiomics-KRAS joint model was constructed by logistic regression, and a nomogram was drawn. The application efficiency of the model was evaluated by the receiver operating characteristic curve and calibration curve. Results:There was a statistically significant difference in KRAS mutation between the pCR group and the non-pCR group in the training set ( χ2=4.69, P=0.032). Ten radiomics features were screened out in MRI images to establish the radiomics model. In the training set and validation set, the areas under the curve (AUC) of KRAS mutation, radiomics model and radiomics-KRAS nomogram model for evaluating pCR after neoadjuvant chemoradiotherapy were 0.665 (95% CI 0.592-0.757), 0.746 (95% CI 0.651-0.895) and 0.818 (95% CI 0.742-0.934), respectively, and the AUCs of the validation set were 0.613 (95% CI 0.582-0.755), 0.738 (95% CI 0.627-0.839) and 0.833 (95% CI 0.768-0.961), respectively. The results of DeLong test showed that the AUC of radiomics-KRAS nomogram model was higher than that of KRAS mutation and radiomics model, and the difference was statistically significant ( Z=0.58, 0.63, P=0.024, 0.022 in the training set; Z=0.54, 0.61, P=0.018, 0.035 in the validation set). The calibration curve showed that the predicted probability of the radiomics-KRAS nomogram model was consistent with the actual probability. Conclusions:The multi-sequence MRI radiomics combined with the KRAS mutation nomogram model has the best efficacy in predicting pCR in patients with rectal cancer after neoadjuvant chemoradiotherapy, and has good practical application value.
