Scholarlink

Purpose <p>The burden of postoperative liver metastasis (LM) in colorectal cancer remains substantial. Consequently, there is a pressing demand for highly effective predictive biomarkers to expedite early diagnosis and treatment.</p> Materials and methods <p>Using a retrospective design, 273 patients were recruited from Xijing Hospital between October 2015 and July 2021. The cohort comprised 91 patients with LM and 182 without LM. The regions of interest in the liver, spleen, and tumor were delineated using portal venous phase CT scans, followed by extraction of radiomic features. Subsequently, radiomics score (rad-score) was developed based on the optimal features identified in each respective region. Finally, a combined predictive model was developed incorporating rad-score and clinicopathological characteristics.</p> Results <p>Multivariable analysis confirmed that the liver, spleen, and tumor radiomics scores were each independent predictors of LM. The combined radiomic–clinical model demonstrated strong discriminative ability, with AUCs of 0.866 in the training set and 0.814 in the validation set, and a nomogram was subsequently developed. The calibration curves indicated a strong degree of agreement between the predicted values and the actual event probabilities. Decision curve analysis demonstrated that the nomogram could provide a substantial net benefit, indicating that its use would result in a considerable improvement in clinical outcomes across a broad range of situations.</p> Conclusions <p>The combined model exhibited satisfactory predictive performance, thereby contributing to the enhancement of clinical diagnostic accuracy and prognostic predictions for LM in CRC patients.</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

Predicting postoperative liver metastasis in colorectal cancer through CT radiomics model based on liver, spleen, and tumor

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月20日

查看全文
  • Xinyu Dou,
  • Jinsong Zhang,
  • Gaozan Zheng,
  • Zhenyu Xie,
  • Hanjun Dan,
  • Guangming Ren,
  • Ye Tian,
  • Minwen Zheng,
  • Jianyong Zheng,
  • Fan Feng

摘要

Purpose

The burden of postoperative liver metastasis (LM) in colorectal cancer remains substantial. Consequently, there is a pressing demand for highly effective predictive biomarkers to expedite early diagnosis and treatment.

Materials and methods

Using a retrospective design, 273 patients were recruited from Xijing Hospital between October 2015 and July 2021. The cohort comprised 91 patients with LM and 182 without LM. The regions of interest in the liver, spleen, and tumor were delineated using portal venous phase CT scans, followed by extraction of radiomic features. Subsequently, radiomics score (rad-score) was developed based on the optimal features identified in each respective region. Finally, a combined predictive model was developed incorporating rad-score and clinicopathological characteristics.

Results

Multivariable analysis confirmed that the liver, spleen, and tumor radiomics scores were each independent predictors of LM. The combined radiomic–clinical model demonstrated strong discriminative ability, with AUCs of 0.866 in the training set and 0.814 in the validation set, and a nomogram was subsequently developed. The calibration curves indicated a strong degree of agreement between the predicted values and the actual event probabilities. Decision curve analysis demonstrated that the nomogram could provide a substantial net benefit, indicating that its use would result in a considerable improvement in clinical outcomes across a broad range of situations.

Conclusions

The combined model exhibited satisfactory predictive performance, thereby contributing to the enhancement of clinical diagnostic accuracy and prognostic predictions for LM in CRC patients.