基于多参数磁共振成像的影像组学列线图在识别脑白质高信号半暗带中的应用研究

Application of Radiomics Nomogram Based on Multi-Parameter of Magnetic Resonance Imaging for Detect the Penumbra of White Matter Hyperintensity

目的探讨使用T2WI联合FLAIR序列成像的影像组学列线图识别脑白质高信号半暗带的临床可行性。方法回顾性分析2012年至2017年我院收治的57例白质高信号进展患者和57例健康体检者的资料并分为训练组(n=80)和验证组(n=34)。对每例患者和健康体检的白质高信号半暗带区分别对T2WI和FLAIR图像提取纹理特征,然后对训练组使用最小绝对收缩与选择算子算法进行降维并建立影像组学标签,同时结合临床危险因素作为Logistic回归分析的潜在预测因子构建识预测模型并制作模型列线图。通过验证组数据评估影像组学列线图的性能和临床应用价值。结果列线图预测白质高信号半暗带的准确率在训练组和验证组中分别为0.881和0.862。Hosmer-Lemeshow检验在训练和验证集间差异均无统计学意义(P>0.05)。使用受试者工作特征曲线评估列线图、影像组学标签和糖尿病因素的AUC分别为0.829、0.819和0.64。结论基于多参数磁共振成像的影像组学列线图可作为一种量化工具帮助临床医生识别白质高信号半暗带,从而进行早期的干预治疗。

Objective To explore the clinical feasibility of detecting the penumbra of white matter hyperintensity(WMHp)by radiomics nomogram based on T2 WI and FLAIR MR images.Methods we performed a retrospective analysis of 57 patients with white matter high signal progression and 57 health examinees from 2012 to 2017 in our hospital.All of the subjects were then divided into training set(n=80)and validation set(n=34).Texture features were extracted from T2 WI and FLAIR images in the WMHp of each patient and healthy subject,respectively.Radiomics signature was established by dimensionally reducing features in the training set by the least absolute shrinkage and selection operator,and the prediction model was constructed with clinical risk factors as potential predictors of Logistic regression analysis,then the nomogram was made.Performance of nomogram and clinical application value was evaluated by data of validation set.Results The accuracy of nomogram to predict WMHp was 0.881 and 0.862 in the training and validation sets,respectively.Hosmer-lemeshow test showed no statistically significant difference between training and validation sets(P>0.05).Receiver opercting characteristic curves were used to evaluate the accuracy of the nomogram,radiomics signature and diabetes factors,with the AUC of 0.829,0.819 and 0.64,respectively.Conclusion The radiomics monogram based on multi-parameter MR images can be used as a quantitative tool to help clinicians identify WMHp for early intervention treatment.