Scholarlink

Objective <p>To investigate the clinical value of a nomogram integrating placental MR radiomics features, MR sign scores, and clinical indicators for prenatal diagnosis of placenta accreta spectrum disorders (PAS) and risk assessment of adverse clinical outcomes.</p> Methods <p>This retrospective study analyzed clinical and imaging data from 167 pregnant women (89 PAS cases, 78 non-PAS) randomly allocated into training (<i>n</i> = 119) and validation (<i>n</i> = 48) sets. Subjective MR signs were evaluated to establish an MR scoring system. Radiomics features were extracted from two sequences. The least absolute shrinkage and selection operator (LASSO) algorithm selected features to construct a radiomics model, generating a radiomics score (Radscore). Multivariate logistic regression combined clinical indicators, MR scores, and Radscore to develop joint prediction models. Model performance was assessed using ROC curves, bootstrap validation, and decision curve analysis. The optimal model was visualized as a nomogram (Nomoscore), which was further evaluated for PAS subtype differentiation and adverse outcome prediction.</p> Results <p>Among all models, the combined model incorporating induced abortion number, MR scores, and Radscore demonstrated superior diagnostic performance, achieving AUCs of 0.857 (95% <i>CI</i>: 0.792 ~ 0.922) and 0.848 (95% <i>CI</i>: 0.74 ~ 0.956) in the training and validation sets, respectively. At a threshold &gt; 0.452, this outperformed standalone models (MR score, clinical model, clinical-MR score model, radiomics model; <i>Z</i> values were 2.764, 3.218, 2.470, 2.213, all <i>p</i> &lt; 0.05). The nomogram effectively differentiated placenta accreta (PA) from placenta increta (PI) (AUC = 0.837, 95% <i>CI</i>: 0.769 ~ 0.905) and PI from placenta percreta (PP) (AUC = 0.879, 95% <i>CI</i>: 0.807 ~ 0.951), while predicting adverse outcomes with an AUC of 0.822 (95% <i>CI</i>: 0.753 ~ 0.891) at optimal thresholds (&gt; 0.628, &gt; 0.966, and &gt; 0.710, respectively). Decision curve analysis confirmed higher clinical net benefit compared to alternative models.</p> Conclusion <p>The nomogram integrating MR radiomics, MR signs, and induced abortion number offers potential clinical utility for prenatal PAS diagnosis, subtype classification, and risk stratification of adverse outcomes.</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

A nomogram based on MR radiomics and MR sign score for prenatal diagnosis of placenta accreta spectrum disorders and risk assessment of adverse clinical outcomes

  • 期刊论文

  • 发表日期：

    2025年8月13日

查看全文
  • Mengzhe Zuo,
  • Qian Chu,
  • Yi Zhang,
  • Zhen Zhang,
  • Ting Pan,
  • Chunlei Zhang,
  • Qin Wang

摘要

Objective

To investigate the clinical value of a nomogram integrating placental MR radiomics features, MR sign scores, and clinical indicators for prenatal diagnosis of placenta accreta spectrum disorders (PAS) and risk assessment of adverse clinical outcomes.

Methods

This retrospective study analyzed clinical and imaging data from 167 pregnant women (89 PAS cases, 78 non-PAS) randomly allocated into training (n = 119) and validation (n = 48) sets. Subjective MR signs were evaluated to establish an MR scoring system. Radiomics features were extracted from two sequences. The least absolute shrinkage and selection operator (LASSO) algorithm selected features to construct a radiomics model, generating a radiomics score (Radscore). Multivariate logistic regression combined clinical indicators, MR scores, and Radscore to develop joint prediction models. Model performance was assessed using ROC curves, bootstrap validation, and decision curve analysis. The optimal model was visualized as a nomogram (Nomoscore), which was further evaluated for PAS subtype differentiation and adverse outcome prediction.

Results

Among all models, the combined model incorporating induced abortion number, MR scores, and Radscore demonstrated superior diagnostic performance, achieving AUCs of 0.857 (95% CI: 0.792 ~ 0.922) and 0.848 (95% CI: 0.74 ~ 0.956) in the training and validation sets, respectively. At a threshold > 0.452, this outperformed standalone models (MR score, clinical model, clinical-MR score model, radiomics model; Z values were 2.764, 3.218, 2.470, 2.213, all p < 0.05). The nomogram effectively differentiated placenta accreta (PA) from placenta increta (PI) (AUC = 0.837, 95% CI: 0.769 ~ 0.905) and PI from placenta percreta (PP) (AUC = 0.879, 95% CI: 0.807 ~ 0.951), while predicting adverse outcomes with an AUC of 0.822 (95% CI: 0.753 ~ 0.891) at optimal thresholds (> 0.628, > 0.966, and > 0.710, respectively). Decision curve analysis confirmed higher clinical net benefit compared to alternative models.

Conclusion

The nomogram integrating MR radiomics, MR signs, and induced abortion number offers potential clinical utility for prenatal PAS diagnosis, subtype classification, and risk stratification of adverse outcomes.