Scholarlink

Purpose <p>This study aimed to develop a nomogram combining clinical, sonographic, and radiomic features to discriminate invasive breast cancer (IBC) from noninvasive breast cancer (non-IBC), and to evaluate the prognostic potential&#xa0;of conventional ultrasound (CUS)-based radiomic signatures in predicting breast cancer invasiveness.</p> Methods <p>A total of 403 IBCs and 221 non-IBCs were retrospectively collected from multiple institutes. The cases were divided into three subsets based on their institutional origin: a training set (<i>n</i> = 353), an internal test set (<i>n</i> = 153), and an external test set (<i>n</i> = 118). A total of 1125 radiomic features were extracted from the training set of CUS images, and Radiomics Scores (Rad-scores) were constructed using the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm. Different nomogram models were constructed using logistic regression, including a clinical–radiomics model (Clinic + Rad), a CUS–clinical model (CUS + Clinic), and a combined CUS–clinical–radiomics model (CUS + Clinic + Rad). The diagnostic performances of these different models were assessed and compared by calculating the area under the receiver operating curve (AUC) as well as the corresponding sensitivity and specificity from the internal and external test sets.</p> Results <p>Significant differences were observed between non-IBC and IBC groups in the following variables:&#xa0;Rad-score, age, axillary lymph node metastasis (ALNM), speculated margin, and blood flow (all <i>P</i> &lt; 0.05). On the basis of these factors, the CUS + Clinic + Rad model significantly outperformed other models, with AUC values of 0.91 in the training set, 0.94 in the internal test set, and 0.90 in the external test set(all <i>P</i> &lt; 0.05). Furthermore, the combined model demonstrated significantly higher sensitivity compared to the single Rad-score model&#xa0;(91.7% vs. 80.0%, <i>P</i> &lt; 0.05), while no significant difference was observed in specificity (83.7% vs. 79.6%, <i>P</i> &gt; 0.05). The proposed combined nomogram demonstrated excellent calibration&#xa0;and clinical utility.</p> Conclusions <p>Radiomic features significantly enhanced radiologists' diagnostic accuracy in distinguishing non-IBC from IBC. The combined CUS–clinical–radiomics model showed robust performance in predicting invasive status of breast cancer, highlighting its potential for clinical translation.</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

Ultrasound-based radiomic nomogram for predicting the invasive status of breast cancer: a multicenter study

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月1日

查看全文
  • Dan Yan,
  • Jingwen Xie,
  • Wanling Cheng,
  • Wen Xue,
  • Yaohong Den,
  • JianXing Zhang

摘要

Purpose

This study aimed to develop a nomogram combining clinical, sonographic, and radiomic features to discriminate invasive breast cancer (IBC) from noninvasive breast cancer (non-IBC), and to evaluate the prognostic potential of conventional ultrasound (CUS)-based radiomic signatures in predicting breast cancer invasiveness.

Methods

A total of 403 IBCs and 221 non-IBCs were retrospectively collected from multiple institutes. The cases were divided into three subsets based on their institutional origin: a training set (n = 353), an internal test set (n = 153), and an external test set (n = 118). A total of 1125 radiomic features were extracted from the training set of CUS images, and Radiomics Scores (Rad-scores) were constructed using the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm. Different nomogram models were constructed using logistic regression, including a clinical–radiomics model (Clinic + Rad), a CUS–clinical model (CUS + Clinic), and a combined CUS–clinical–radiomics model (CUS + Clinic + Rad). The diagnostic performances of these different models were assessed and compared by calculating the area under the receiver operating curve (AUC) as well as the corresponding sensitivity and specificity from the internal and external test sets.

Results

Significant differences were observed between non-IBC and IBC groups in the following variables: Rad-score, age, axillary lymph node metastasis (ALNM), speculated margin, and blood flow (all P < 0.05). On the basis of these factors, the CUS + Clinic + Rad model significantly outperformed other models, with AUC values of 0.91 in the training set, 0.94 in the internal test set, and 0.90 in the external test set(all P < 0.05). Furthermore, the combined model demonstrated significantly higher sensitivity compared to the single Rad-score model (91.7% vs. 80.0%, P < 0.05), while no significant difference was observed in specificity (83.7% vs. 79.6%, P > 0.05). The proposed combined nomogram demonstrated excellent calibration and clinical utility.

Conclusions

Radiomic features significantly enhanced radiologists' diagnostic accuracy in distinguishing non-IBC from IBC. The combined CUS–clinical–radiomics model showed robust performance in predicting invasive status of breast cancer, highlighting its potential for clinical translation.