Texture Analysis of Sequential Images of T2-weighted Imaging and Diffusion-weighted Imaging for Predicting the Efficacy of Chemoradiotherapy in Cervical Squamous Cell Carcinoma

Abstract

Objective To investigate the correlation of the texture parameters of T2-weighted imaging (T2WI) and diffusion-weighted imaging (DWI) with the efficacy of chemoradiotherapy in cervical squamous cell carcinoma. Methods Patients with squamous cell carcinoma of the cervix that underwent chemoradiotherapy from February 2015 to January 2016 in Peking Union Medical College Hospital were included retrospectively, and were divided into the disease-progressive group and the disease-stable group according to their prognosis. Texture analysis of baseline T2WI and DWI images before chemoradiotherapy was carried out with Texrad software, and the texture parameters of spatial scale filter (SSF) with radius values of 2, 4 and 6 were obtained. The differences of texture parameters between the two groups were compared, and the correlation between the texture parameters and the curative of chemoradiotherapy in patients with cervical squamous cell carcinoma was analyzed by multivariate Cox regression. Receiver operating characteristic (ROC) curve was used to analyze the performance of texture parameters in predicting disease progression after chemoradiotherapy in patients with cervical squamous cell carcinoma. Results A total of 121 patients with squamous cell carcinoma of the cervix that met the inclusion and exclusion criteria were enrolled in this study. There were 46 cases in the disease-progressive group and 75 cases in the disease-stable group. In T2WI sequential images, there were significant differences in the texture parameters of means (SSF2, SSF4, SSF6), skewness (SSF2, SSF4), and entropy (SSF4, SSF6) between disease-progressive group and disease-stable group (all P < 0.05). In DWI sequential images, there were significant differences in the texture parameters of means (SSF2, SSF4, SSF6), skewness (SSF4, SSF6), and kurtosis (SSF2, SSF4) between the two groups (all P < 0.05). Multivariate Cox regression analysis showed that the texture parameter of means (SSF2, SSF4, SSF6) of T2WI and the texture parameters of means (SSF2, SSF6), entropy (SSF2, SSF4, SSF6) and skewness (SSF4, SSF6) of DWI were correlated with the efficacy of chemoradiotherapy in patients with cervical squamous cell carcinoma (P < 0.05). The Results of ROC analysis showed that the texture parameter of means (T2WI-SSF2, T2WI-SSF4, T2WI-SSF6, DWI-SSF2, DWI-SSF6) and skewness (DWI-SSF6) could predict the progression of cervical squamous cell carcinoma after chemoradiotherapy in patients with cervical squamous cell carcinoma. The area under the curve (AUC) was 0.625-0.746. Among them, the mean of T2WI-SSF4 was the most effective (AUC: 0.746), followed by the mean of T2WI-SSF2 (AUC: 0.725) and the mean of T2WI-SSF6 (AUC: 0.703). Conclusions The texture parameters of baseline T2WI and DWI sequences were correlated with the curative effect of chemoradiotherapy in patients with cervical squamous cell carcinoma. The parameters of means and skewness can predict the progression of cervical squamous carcinoma after chemoradiotherapy, and the mean has a higher predictive power.

Keywords

• cervical carcinoma
• image texture analysis
• t2-weighted imaging
• diffusion-weighted imaging
• chemoradiotherapy