| 李西瑞,王得志,杨晓楠,李杰,郝大鹏,崔久法.基于增强MRI影像组学、深度学习及临床特征构建列线图模型鉴别脊柱结核与化脓性脊柱炎[J].中国医学影像技术,2025,41(1):122~127 |
| 基于增强MRI影像组学、深度学习及临床特征构建列线图模型鉴别脊柱结核与化脓性脊柱炎 |
| Nomogram model based on enhanced MRI radiomics, deep learning and clinical features for differentiating spinal tuberculosis and pyogenic spondylitis |
| 投稿时间：2024-06-30 修订日期：2024-10-14 |
| DOI：10.13929/j.issn.1003-3289.2025.01.026 |
| 中文关键词: 结核,脊柱 脊柱炎 磁共振成像 深度学习 影像组学 |
| 英文关键词:tuberculosis, spinal spondylitis magnetic resonance imaging deep learning radiomics |
| 基金项目: |
|
| 摘要点击次数: 395 |
| 全文下载次数: 128 |
| 中文摘要: |
| 目的 评价基于增强MRI影像组学、深度学习(DL)及临床特征建立的列线图模型用于鉴别脊柱结核与化脓性脊柱炎的价值。方法 回顾性纳入59例脊柱结核、66例化脓性脊柱炎,筛选可用于鉴别脊柱结核和化脓性脊柱炎的影像组学、DL及临床特征;以logistic回归基于最优特征构建预测模型,并联合以上特征构建列线图模型。以受试者工作特征曲线、校准曲线和决策曲线可视化模型鉴别脊柱结核与化脓性脊柱炎的效能。结果 列线图鉴别训练集和测试集脊柱结核与化脓性脊柱炎的曲线下面积(AUC)均最高,分别为0.997和0.920。DeLong检验显示列线图模型与临床模型在测试集的AUC差异有统计学意义(P=0.002),而与其他模型差异均无统计学意义(P均＞0.05)。列线图模型可为鉴别脊柱结核与化脓性脊柱炎提供最高的总净获益,且其校准度良好。结论 基于增强MRI影像组学、DL及临床特征构建的列线图模型用于鉴别脊柱结核与化脓性脊柱炎具有较高效能。 |
| 英文摘要: |
| Objective To observe the efficacy of nomogram model based on enhanced MRI radiomics, deep learning (DL) and clinical features for differentiating spinal tuberculosis and pyogenic spondylitis. Methods Totally 59 cases of spinal tuberculosis and 66 of pyogenic spondylitis were retrospectively enrolled. Radiomics, DL and clinical features relevant to differentiating spinal tuberculosis and pyogenic spondylitis were selected. Then a predictive model was constructed using logistic regression based on the selected optimal features, and a comprehensive nomogram model was developed through combination of the above features. The effectiveness of these models for distinguishing spinal tuberculosis from pyogenic spondylitis were visualized based on receiver operating characteristic curves, calidration curves and decision curves. Results The nomogram model demonstrated the highest area under the curve (AUC) in both training set and test set, with AUC of 0.997 and 0.920, respectively. In test set, DeLong test indicated that the difference of AUC between the nomogram model and clinical model was significant (P=0.002), while no significant difference was observed between the nomogram model and the other models (all P＞0.05). The nomogram model provided the highest overall net benefit and exhibited good calibration for distinguishing spinal tuberculosis from pyogenic spondylitis. Conclusion Nomogram model based on enhanced MRI radiomics, DL and clinical features demonstrated high efficacy for differentiating spinal tuberculosis from pyogenic spondylitis. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
