基于MRI影像组学模型鉴别鼻腔鼻窦内翻乳头状瘤与息肉

Differential diagnosis of inverted papilloma and polyp of nasal cavity and paranasal sinus based on MRI imaging model

目的探讨基于磁共振T2WI及增强T1WI影像组学模型在鉴别鼻腔鼻窦内翻乳头状瘤与息肉的应用价值。方法搜集病理证实且具有完整T2WI及增强T1WI图像的内翻乳头状瘤(NIP)病例54例,单侧鼻息肉(NP)病例51例。采用ITK-Snap勾画病灶的所有层面。采用pyradiomics提取影像组学特征,首先使用最小冗余最大相关(mRMR)进行特征提取,进一步使用rfe-SVM特征消除,去掉最小得分的特征,建立预测模型。使用ROC曲线敏感性和特异性评价模型的性能,并在验证集中验证。结果总共提取影像组学特征1133个,经mRMR降维后筛选保留30个特征用于建立预测模型。T1WI增强预测模型的AUC值为:训练集0.98,验证集0.95,训练集的敏感度和特异度分别为89.7%和100%,验证集的敏感度和特异度分别为93.8%和93.3%。T2WI预测模型的AUC值为:训练集0.95,验证集0.91,训练集的敏感度和特异度分别为82.1%和95.6%,验证集的敏感度和特异度分别为93.8%和84.2%。结论基于磁共振影像组学T1WI增强预测模型及T2WI预测模型对鉴别鼻腔鼻窦内翻乳头状瘤及息肉有一定的价值。

Objective To explore the value of T2WI and enhanced T1WI magnetic resonance imaging models in differentiating inverted papilloma and polyp of nasal cavity and paranasal sinus.Methods 54 cases of inverted papilloma(NIP)and 51 cases of polyp(NP)with complete T2WI and enhanced T1WI images confirmed by pathology were collected.ITK snap was used to outline all levels of the lesion.Pyromics was used to extract image omics features.Firstly,the(mRMR)was used for feature extraction,and then rfe-SVM was used to remove the feature of minimum score and establish a prediction model.The sensitivity and specificity of ROC curve were used to evaluate the performance of the model,and were verified in the validation set.Results A total of 1133 image omics features were extracted,and 30 features were retained after mrmr dimensionality reduction for the establishment of prediction models.The AUC value of T1WI enhanced prediction model was:training set 0.98,validation set 0.95,the sensitivity and specificity of training set were 89.7% and 100% respectively,and the sensitivity and specificity of validation set were 93.8% and 93.3% respectively.The AUC value of T2WI prediction model was:training set 0.95,validation set 0.91,the sensitivity and specificity of training set were 82.1% and 95.6% respectively,and the sensitivity and specificity of validation set were 93.8% and 84.2% respectively.Conclusion MRI based on T1WI enhancement prediction model and T2WI prediction model have certain value in differentiating inverted papilloma and polyp of nasal cavity and paranasal sinus.