Multi-sequence MRI texture analysis for predicting efficacy of neoadjuvant chemotherapy in uterine cervical carcinoma / 中国医学影像技术
Chinese Journal of Medical Imaging Technology
; (12): 1215-1219, 2020.
Article
em Zh
| WPRIM
| ID: wpr-860943
Biblioteca responsável:
WPRO
ABSTRACT
Objective: To observe the value of multi-sequence MRI texture analysis for predicting efficacy of neoadjuvant chemotherapy (NACT) in uterine cervical carcinoma. Methods: A total of 32 cervical carcinoma patients underwent NACT. Pelvic MRI was performed before and after NACT, and the patients were divided into response group (complete response and partial response) and non-response group (stable disease and progressive disease) according to the standards of response evaluation criteria in solid tumors (RECIST). Texture analysis was performed on T2WI, DWI and contrast enhanced (CE) images before NACT. Totally 106 parameters were obtained for each sequence and compared between the two groups. Multivariate Logistic regression analysis was performed using the top two significant parameters as independent variables and create formula. By drawing the receiver operating characteristic (ROC) curve, the predictive value of single factor and regression model of each sequence were obtained, and the comparison was made among sequences. Results: Significant differences of 22 texture features of T2WI, 13 of DWI and 36 of CE were found between response group and non-response group (all P<0.05), and the areas under curve (AUC) using single features for predicting the effect of NACT for cervical cancer were 0.609-0.839, 0.745-0.813 and 0.552-0.786, respectively. Multivariate Logistic regression analysis for each sequence showed AUC of 0.839 for T2WI, 0.885 for DWI and 0.766 for CE. Conclusion: Pre-NACT multi-sequence MRI texture analysis has potential in predicting efficacy of NACT in cervical carcinoma, and DWI may be the best.
Texto completo:
1
Índice:
WPRIM
Tipo de estudo:
Prognostic_studies
Idioma:
Zh
Revista:
Chinese Journal of Medical Imaging Technology
Ano de publicação:
2020
Tipo de documento:
Article