Scholarlink

Background <p>The objective of this study was to explore the association between radiomic features changes extracted from MRI before and after concurrent chemoradiotherapy (CCRT) and disease recurrence, as well as disease-free survival, in patients with locally advanced cervical cancer.</p> Materials and methods <p>Thirty-seven patients with International Federation of Gynecology and Obstetrics stage IIB–IVA cervical cancer were included in this retrospective study. All patients underwent pelvic MRI before and after CCRT. Texture features were extracted using LIFEx software. Regions of interest were manually segmented on T2-weighted images and apparent diffusion coefficient&#xa0;(ADC) maps by a junior radiologist under the supervision of a senior radiologist with 15&#xa0;years of experience. In post-treatment scans without visible residual tumor, regions of interest were drawn over the cervix. Delta-radiomic features were calculated using the following equation: (post-CCRT − pre-CCRT)/pre-CCRT × 100. Disease recurrence was defined as the primary endpoint, while disease-free survival and machine learning analyses were performed for exploratory and supportive purposes. Statistical analyses included the Wilcoxon–Mann–Whitney test, receiver operating characteristic analysis, Kaplan–Meier method with log-rank test, and Cox regression analysis. Machine learning performance for recurrence classification was evaluated using Orange Data Mining software.</p> Results <p>In multivariate analysis, the Gray-Level Size Zone Matrix–High Gray&#xa0;Level Zone Emphasis feature derived from the pre-treatment ADC map was independently associated with disease-free survival. In machine learning analysis, the highest performance for recurrence classification was achieved using delta-radiomic features with the random forest algorithm, yielding an area under the curve of 0.897.</p> Conclusion <p>Radiomic analysis of pre- and post-treatment MRI scans shows potential as a noninvasive tool for the assessment recurrence in patients with locally advanced cervical cancer.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Delta-radiomic features on diffusion-weighted MRI before and after concurrent chemoradiotherapy in locally aggressive cervical cancer: an exploratory study

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月8日

查看全文
  • Ozge Coskun,
  • Hilal Sahin,
  • Zeliha Guzeloz Capar

摘要

Background

The objective of this study was to explore the association between radiomic features changes extracted from MRI before and after concurrent chemoradiotherapy (CCRT) and disease recurrence, as well as disease-free survival, in patients with locally advanced cervical cancer.

Materials and methods

Thirty-seven patients with International Federation of Gynecology and Obstetrics stage IIB–IVA cervical cancer were included in this retrospective study. All patients underwent pelvic MRI before and after CCRT. Texture features were extracted using LIFEx software. Regions of interest were manually segmented on T2-weighted images and apparent diffusion coefficient (ADC) maps by a junior radiologist under the supervision of a senior radiologist with 15 years of experience. In post-treatment scans without visible residual tumor, regions of interest were drawn over the cervix. Delta-radiomic features were calculated using the following equation: (post-CCRT − pre-CCRT)/pre-CCRT × 100. Disease recurrence was defined as the primary endpoint, while disease-free survival and machine learning analyses were performed for exploratory and supportive purposes. Statistical analyses included the Wilcoxon–Mann–Whitney test, receiver operating characteristic analysis, Kaplan–Meier method with log-rank test, and Cox regression analysis. Machine learning performance for recurrence classification was evaluated using Orange Data Mining software.

Results

In multivariate analysis, the Gray-Level Size Zone Matrix–High Gray Level Zone Emphasis feature derived from the pre-treatment ADC map was independently associated with disease-free survival. In machine learning analysis, the highest performance for recurrence classification was achieved using delta-radiomic features with the random forest algorithm, yielding an area under the curve of 0.897.

Conclusion

Radiomic analysis of pre- and post-treatment MRI scans shows potential as a noninvasive tool for the assessment recurrence in patients with locally advanced cervical cancer.