Department of Radiology, Qionglai Medical Center Hospital, No. 172 Xinglin Road, Wenjun Street, Qionglai, Sichuan, 611530, People's Republic of China (Q.F., T.W.).
Department of Radiology, Luzhou Traditional Chinese Medicine Hospital, No. 11 Jiangyang South Road, Luzhou, Sichuan, 646000, People's Republic of China (S.X.).
Acad Radiol. 2024 Feb;31(2):605-616. doi: 10.1016/j.acra.2023.07.011. Epub 2023 Aug 14.
This study aimed to develop and validate a magnetic resonance imaging (MRI)-based radiomics nomogram combining radiomics signatures and clinical factors to differentiate between benign and malignant vertebral compression fractures (VCFs).
A total of 189 patients with benign VCFs (n = 112) or malignant VCFs (n = 77) were divided into training (n = 133) and validation (n = 56) cohorts. Radiomics features were extracted from MRI T1-weighted images and short-TI inversion recovery images to develop the radiomics signature, and the Rad score was constructed using least absolute shrinkage and selection operator regression. Demographic and MRI morphological characteristics were assessed to build a clinical factor model using multivariate logistic regression analysis. A radiomics nomogram was constructed based on the Rad score and independent clinical factors. Finally, the diagnostic performance of the radiomics nomogram, clinical model, and radiomics signature was validated using receiver operating characteristic and decision curve analysis (DCA).
Six features were used to build a combined radiomics model (combined-RS). Pedicle or posterior element involvement, paraspinal mass, and fluid sign were identified as the most important morphological factors for building the clinical factor model. The radiomics signature was superior to the clinical model in terms of the area under the curve (AUC), accuracy, and specificity. The radiomics nomogram integrating the combined-RS, pedicle or posterior element involvement, paraspinal mass, and fluid sign achieved favorable predictive efficacy, generating AUCs of 0.92 and 0.90 in the training and validation cohorts, respectively. The DCA indicated good clinical usefulness of the radiomics nomogram.
The MRI-based radiomics nomogram, combining the radiomics signature and clinical factors, showed favorable predictive efficacy for differentiating benign from malignant VCFs.
本研究旨在开发并验证一种基于磁共振成像（MRI）的放射组学列线图，该列线图结合放射组学特征和临床因素，用于区分良性和恶性椎体压缩性骨折（VCF）。
共纳入 189 例良性 VCF（n=112）或恶性 VCF（n=77）患者，分为训练集（n=133）和验证集（n=56）。从 MRI T1 加权图像和短 TI 反转恢复图像中提取放射组学特征，使用最小绝对收缩和选择算子回归构建 Rad 评分。使用多变量逻辑回归分析评估人口统计学和 MRI 形态学特征，以构建临床因素模型。基于 Rad 评分和独立的临床因素构建放射组学列线图。最后，使用受试者工作特征和决策曲线分析（DCA）验证放射组学列线图、临床模型和放射组学特征的诊断性能。
共构建了 6 个特征的联合放射组学模型（combined-RS）。椎弓根或后弓累及、椎旁肿块和液性征被确定为构建临床因素模型的最重要形态学因素。在曲线下面积（AUC）、准确性和特异性方面，放射组学特征均优于临床模型。整合联合-RS、椎弓根或后弓累及、椎旁肿块和液性征的放射组学列线图具有良好的预测效能，在训练集和验证集中的 AUC 分别为 0.92 和 0.90。DCA 表明放射组学列线图具有良好的临床实用性。
基于 MRI 的放射组学列线图，结合放射组学特征和临床因素，对区分良性和恶性 VCF 具有良好的预测效能。
