多模态MRI影像组学评估前列腺癌的侵袭性

Radiomics Based on Multiparametric MRI:Assessing the Aggressiveness of Prostate Cancer

目的探讨基于多参数磁共振(mp-MRI)的影像组学模型在评估前列腺癌侵袭性中的应用价值。方法回顾性纳入120例行腹腔镜下前列腺癌根治切除术的患者,并根据病理结果将患者分为低危组[(GS)≤3+4,共40例]和高危组(GS≥4+3,共80例)。将患者按7∶3的比例分配到训练组及验证组。从患者T2WI、DWI、DCE图像上提取肿瘤区域的组学特征,通过LASSO回归选取用于建模的特征,同时应用多因素逻辑回归来联合组学特征和临床特征,建立术前鉴别低危组和高危组PCa的联合模型。使用受试者工作特征(ROC)曲线分析模型的诊断价值,并在验证组中对诊断效能进行验证。结果训练组共纳入84例,包含29个低危和55个高危PCa;验证组纳入36例,包含11个低危和25个高危PCa。T2WI、DWI、DCE每个序列均提取1304个组学特征,经过LASSO回归后,仅4个组学特征被纳入组学模型;通过结合PSA,建立联合模型,其在训练组的AUC为0.846,在验证组中为0.800。结论基于多参数磁共振的影像组学可用于术前鉴别低危及高危组PCa。

Objective To develop and validate a combined model based on multiparametric MRI(mp-MRI)for assessing the aggressiveness of prostate cancer(PCa).Methods This was a retrospective analysis of 120 patients treated with laparoscopic radical prostatectomy.According to the pathology,these patients were classified as the low-risk(GS≤3+4,40 patients)and high-risk PCa(GS≥4+3,80 patients).Then,these patients were allocated to a training or validation set,with a ratio of 7:3.Radiomic features were extracted from T2WI,DWI,and DCE images of lesions.The least absolute shrinkage and selection operator(LASSO)regression analysis was used for feature selection.Multivariate logistic regression analysis was used to combine radiomic features with the clinical variables to build the combined model.Receiver operating characteristic(ROC)curves were constructed to evaluate the diagnostic R performance of the models for predicting aggressiveness of PCa in both training and validation groups.esults A total of 84 patients(29 low-risk and 55 high-risk lesions)and 36(11 low-risk and 25 highrisk lesions)were allocated to the training cohort and validation cohort,respectively.Although a large number of 1304 features were extracted from each sequence(T2WI,DWI,or DCE sequence),only 4 radiomic features were included in the model after performing the LASSO method.The combined model,including PSA and radiomic features,was constructed for differentiating the low-risk from high-risk PCa.The area under the curve(AUC)of this model was 0.846 and 0.800 in the training and validation groups.Conclusion The model combining PSA with radiomic features can be used to assess the aggressiveness of PCa.