| 杨钟,付宝月,陈玉兰,李乃玉,方梦诗,孙铭洁,韦超.MRI纹理特征联合表现弥散系数鉴别诊断子宫肉瘤与细胞型子宫肌瘤[J].中国医学影像技术,2024,40(7):1052~1057 |
| MRI纹理特征联合表现弥散系数鉴别诊断子宫肉瘤与细胞型子宫肌瘤 |
| MRI texture features combined with apparent diffusion coefficient for differentiating uterine sarcoma and cellular uterine leiomyoma |
| 投稿时间：2024-01-17 修订日期：2024-02-26 |
| DOI：10.13929/j.issn.1003-3289.2024.07.019 |
| 中文关键词: 子宫肿瘤 肉瘤 平滑肌瘤 磁共振成像 纹理分析 诊断,鉴别 |
| 英文关键词:uterine neoplasms sarcoma leiomyoma magnetic resonance imaging texture analysis diagnosis, differential |
| 基金项目:安徽省教育厅高等学校科研计划(2022AH051262)、安徽省自然科学基金(1908085QH364)。 |
|
| 摘要点击次数: 635 |
| 全文下载次数: 257 |
| 中文摘要: |
| 目的 观察MRI纹理特征联合表观弥散系数(ADC)鉴别诊断子宫肉瘤(US)与细胞型子宫肌瘤(CUL)的价值。方法 回顾性分析27例US(US组)及34例CUL(CUL组)患者盆腔MRI资料,于T2WI及弥散加权成像(DWI)中提取病灶纹理特征,测量ADC值,记录平均ADC值(ADCmean)、最小ADC值(ADCmin)及标准ADC值(ADCst);分别基于ADC值、最优纹理特征及二者联合构建鉴别US与CUL的逻辑回归(LR)模型,包括LRADC、LR纹理及LRADC+纹理模型;绘制受试者工作特征曲线,计算曲线下面积(AUC),评估各模型鉴别诊断US与CUL的效能。结果 US组病灶ADCmean、ADCmin及ADCst均低于CUL组(P均＜0.05)。共基于盆腔T2WI及DWI提取3 750个纹理特征,经筛选后得到5个最优特征,以之构建的LRADC+纹理模型及LR纹理模型鉴别US与CUL的效能相当(AUC=0.921、0.887;P＞0.05),且均高于LRADC模型(AUC=0.696;P均＜0.05)。LRADC+纹理模型的校准曲线与理想曲线走行基本一致,且临床净收益大于LRADC及LR纹理模型。结论 MRI纹理特征联合ADC值可提高鉴别诊断US与CUL的效能。 |
| 英文摘要: |
| Objective To observe the value of MRI texture features combined with apparent diffusion coefficient (ADC) for differentiating uterine sarcoma (US) and cellular uterine leiomyoma (CUL). Methods Pelvic MRI data of 27 US patients (US group) and 34 CUL patients (CUL group) were retrospectively analyzed. The texture features of lesions were extracted from T2WI and diffusion weighted imaging (DWI), the ADC value were measured, and the average ADC value (ADCmean),the minimum ADC value (ADCmin) and standard ADC value (ADCst) were recorded. Then logistic regression (LR) models were constructed based on ADC value, optimal texture features alone and their combination, respectively, including LRADC, LRtexture and LRADC+texture models. Receiver operating characteristic curves were drawn, and the area under the curves (AUC) were calculated to evaluate the efficacy of each model for differentiating US and CUL. Results The ADCmean, ADCmin and ADCst in US group were all lower than those in CUL group (all P＜0.05). A total of 3 750 texture features were extracted from pelvic T2WI and DWI, 5 optimal features were finally obtained, and the constructed LRADC+texture model and LRtexture model had similar efficacy of differentiating US and CUL (AUC=0.921, 0.887; P＞0.05), which were both higher than that of LRADC model (AUC=0.696; both P＜0.05). The calibration curve of LRADC+texture model was basically consistent with the ideal curve, which had better clinical benefits than LRADC and LRtexture models. Conclusion MRI texture features combined with ADC value could improve efficacy for differentiating US and CUL. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
