Scholarlink

Objective <p>To construct a clinical-radiomics nomogram based on T1 weighted imaging (WI) and T2WI of lumbar magnetic resonance imaging (MRI) for predicting osteoporosis.</p> Methods <p>Sixty-eight participants who underwent both dual-energy X-ray absorptiometry and lumbar MRI were included. Participants were classified as having either normal bone mineral density (BMD) (T &gt; − 1) or osteoporosis (T &lt; − 2.5), with those having osteopenia (T-score between − 2.5 and − 1) being excluded. A total of 396 radiomic features (RFs) were extracted from routine lumbar MRI (T1WI and T2WI). Five RFs highly correlated with osteoporosis were selected via logistic regression. The diagnostic values of osteoporosis using a radiomics model and a combined RFs and clinical factors (e.g. age, sex) model were assessed using the receiver operating characteristic (ROC) method. Diagnostic accuracy, sensitivity and specificity were calculated for the clinical-radiomics nomogram.</p> Results <p>In our study of 68 patients (18 men, 50 women), including 33 with osteoporosis and 35 with normal BMD, significant differences were found in age and sex between groups, whereas body mass index was similar. The radiomics model, which analysed 396 features from lumbar MRI, achieved an area under the curve (AUC) of 0.871 (95% CI: 0.768–0.940). Incorporating clinical features into the model improved the AUC to 0.894 (95% CI: 0.796–0.956), with a significant P-value (&lt; 0.0001). Sensitivity increased from 82.86 to 91.43%, whereas specificity decreased from 87.88 to 81.82%. Accuracy rose from 83.8 to 86.8%, and the Akaike information criterion improved from 74.723 to 63.703. Calibration curves indicated good alignment of predicted probabilities with actual outcomes. Decision curve analysis demonstrated enhanced clinical utility for the clinical-radiomics model compared with the radiomics model alone.</p> Conclusion <p>A radiomics model based on routine lumbar MRI can effectively diagnose osteoporosis. The clinical-radiomics nomogram combining RFs with clinical factors improves diagnostic performance.</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

Clinical-radiomics nomogram construction from magnetic resonance imaging to diagnose osteoporosis: a preliminary study

  • 期刊论文

  • 发表日期：

    2025年5月29日

查看全文
  • Ruihong Hou,
  • Wenyun Tan,
  • Chunying Liu,
  • Jiakai Zhang,
  • Shuyan Jing,
  • Li He

摘要

Objective

To construct a clinical-radiomics nomogram based on T1 weighted imaging (WI) and T2WI of lumbar magnetic resonance imaging (MRI) for predicting osteoporosis.

Methods

Sixty-eight participants who underwent both dual-energy X-ray absorptiometry and lumbar MRI were included. Participants were classified as having either normal bone mineral density (BMD) (T > − 1) or osteoporosis (T < − 2.5), with those having osteopenia (T-score between − 2.5 and − 1) being excluded. A total of 396 radiomic features (RFs) were extracted from routine lumbar MRI (T1WI and T2WI). Five RFs highly correlated with osteoporosis were selected via logistic regression. The diagnostic values of osteoporosis using a radiomics model and a combined RFs and clinical factors (e.g. age, sex) model were assessed using the receiver operating characteristic (ROC) method. Diagnostic accuracy, sensitivity and specificity were calculated for the clinical-radiomics nomogram.

Results

In our study of 68 patients (18 men, 50 women), including 33 with osteoporosis and 35 with normal BMD, significant differences were found in age and sex between groups, whereas body mass index was similar. The radiomics model, which analysed 396 features from lumbar MRI, achieved an area under the curve (AUC) of 0.871 (95% CI: 0.768–0.940). Incorporating clinical features into the model improved the AUC to 0.894 (95% CI: 0.796–0.956), with a significant P-value (< 0.0001). Sensitivity increased from 82.86 to 91.43%, whereas specificity decreased from 87.88 to 81.82%. Accuracy rose from 83.8 to 86.8%, and the Akaike information criterion improved from 74.723 to 63.703. Calibration curves indicated good alignment of predicted probabilities with actual outcomes. Decision curve analysis demonstrated enhanced clinical utility for the clinical-radiomics model compared with the radiomics model alone.

Conclusion

A radiomics model based on routine lumbar MRI can effectively diagnose osteoporosis. The clinical-radiomics nomogram combining RFs with clinical factors improves diagnostic performance.