CT signs and AI parameters predict colorectal cancer neoadjuvant chemotherapy efficacy
10.13491/j.issn.1004-714X.2025.05.014
- VernacularTitle:CT征象及人工智能定量参数预测大肠癌新辅助化疗效果
- Author:
Guobin LAN
1
;
Chuang LIU
1
;
Hao WANG
1
;
Hongyu MA
1
;
Zeliang LI
1
;
Wen CHEN
1
;
Wenqiang ZHANG
1
Author Information
1. Imaging Center, Cangzhou Hospital of Integrated TCM-WM Hebei,Cangzhou 061001, China.
- Publication Type:OriginalArticles
- Keywords:
Colorectal cancer;
Neoadjuvant chemotherapy;
Radiomic;
Artificial
- From:
Chinese Journal of Radiological Health
2025;34(5):713-719
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the value of CT signs and quantitative parameters of artificial intelligence (AI) in predicting the efficacy of neoadjuvant chemotherapy for colorectal cancer. Methods A total of 349 colorectal cancer patients who received neoadjuvant chemotherapy at Cangzhou Hospital of Integrated Traditional Chinese and Western Medicine in Hebei Province from January 2022 to January 2025 were selected and and divided into the effective group (n = 267) and the ineffective group (n = 82) according to the evaluation criteria for the efficacy of solid tumors. Conduct a CT examination and extract AI quantitative parameters from the CT images based on the lesion. The data were analyzed using SPSS21.0 software, Logistic regression was used to screen the influencing factors of ineffective neoadjuvant chemotherapy in patients with colorectal cancer, and separate and combined models of CT signs and AI quantitative parameters were established. The predictive effect of the model was verified by using the ROC curve, calibration curve and decision curve. Results Compared with the effective group, the proportion of regular tumor morphology and the proportion of non-enlarged lymph nodesin the ineffective group were smaller. The tumor volume, peak value and entropy value were larger (P < 0.05). Multivariable analysis showed that irregular shape (OR= 4.216), presence of lymph node enlargement (OR = 8.998), larger tumor volume (OR = 1.109), higher average CT value (OR = 1.120), elevated peak value (OR = 2.528), and increased entropy value (OR = 1.390) were independent risk factors for ineffective neoadjuvant chemotherapy in colorectal cancer (P < 0.05). The areas under the ROC curves of the individual and combined models of CT signs and AI quantitative parameters were 0.777, 0.818, and 0.877, respectively(P < 0.05). The calibration curve showed a Brier score of 0.091. The decision curve showed that the threshold was between 0.10 and 0.85, and the combined model achieved a relatively high net clinical benefit. Conclusion CT signs combined with AI quantitative parameters has a predictive value for the efficacy of neoadjuvant chemotherapy in colorectal cancer. To provide evidence-based basis for clinical screening of the population benefiting from chemotherapy and optimization of treatment strategies.
