双参数MRI纹理分析评估前列腺癌的侵袭性价值
Biparametric MRI texture analysis to assess the aggressiveness of prostate cancer patients
姜明瀚 1颜红柱 2张恩榜 1宋黎涛1
作者信息
- 1. 上海中医药大学附属第七人民医院 医学影像科 上海 200137
- 2. 上海中医药大学附属第七人民医院 病理科 上海 200137
- 折叠
摘要
目的 探讨评估前列腺癌患者基于双参数MRI的纹理参数与Gleason评分(GS)之间的相关性,并评估纹理特征在区分临床显著性前列腺癌(CSPC,GS≥7)和非CSPC的诊断性能.方法 选取 96 例经根治性前列腺切除术或活检证实为前列腺癌的患者,同时分为训练组(55 例)和验证组(41 例).所有患者术前均行 3.0T MR检查,使用专用软件对轴位T2WI和ADC图覆盖整个肿瘤体积进行纹理分析.计算相关系数以评估纹理参数与GS之间的相关性,并对显著参数进行多元回归分析.ROC曲线分析预测CSPC的最佳临界值.结果 在训练组数据集中,ADC图的灰度共生矩阵(GLCM)熵是GS的唯一显著指标(决定系数R2=0.424,P=0.002).ADC图GLCM熵的AUC为 0.841(95%CI 0.726～0.928),准确度、敏感度、特异度分别为 83%、88%、73%.当将阈值取 2.91 应用于验证数据集时,其准确度、灵敏度、特异度分别为 91%、96%、71%.结论 ADC图的GLCM熵与GS相关,其区分CSPC和非CSPC的准确率为 83%.
Abstract
Objective To evaluate the correlation between biparametric MRI texture parameters and Gleason score(GS)in prostate cancer patients,and to evaluate the diagnostic performance of texture features for differentiating clinically significant prostate cancer(CSPC,GS≥7)from non-CSPC.Methods 96 prostate cancer patients confirmed by radical prostatectomy or bi-opsy were divided into training set(n=55)and validation set(n=41).All patients underwent 3T-MRI examination,and texture analysis was performed on axial T2WI and ADC maps covering the entire tumor volume using special software.Correlation coeffi-cients were calculated to evaluate the correlation between texture parameters and GS,and multiple regression analysis was fur-ther performed on significant parameters.ROC curve was used to predict the best critical value for CSPC.Results In the train-ing dataset,the gray level co-occurrence matrix(GLCM)entropy of the ADC map was the only significant indicator of GS(coeffi-cient of determination R2=0.424,P=0.002).The AUC of the GLCM entropy on ADC was 0.841(95%CI=0.726-0.928),with an accuracy of 83%,a sensitivity of 88%,and a specificity of 73%.When a threshold of 2.91 was applied to the validation dataset,the accuracy,sensitivity,and specificity were 91%,96%and 71%,respectively.Conclusion The GLCM entropy on ADC map is correlated with GS,and its accuracy in distinguishing CSPC from non-CSPC is 83%.
关键词
前列腺癌/磁共振成像/纹理分析/Gleason评分Key words
Prostate cancer/Magnetic resonance imaging/Texture analysis/Gleason score引用本文复制引用
出版年
2024
NETL
NSTL
万方数据