| 王祥,李萌,王志芳,黄勇华.基于动态增强MRI影像组学形态学特征识别肿块样非特殊型浸润性乳腺癌导管原位癌成分[J].中国医学影像技术,2024,40(12):1856~1860 |
| 基于动态增强MRI影像组学形态学特征识别肿块样非特殊型浸润性乳腺癌导管原位癌成分 |
| Morphological features of mass-like non-special type of invasive breast cancer based on dynamic contrast-enhanced MRI radiomics for identifying ductal carcinoma in situ component |
| 投稿时间：2024-04-16 修订日期：2024-11-11 |
| DOI：10.13929/j.issn.1003-3289.2024.12.011 |
| 中文关键词: 癌,导管,乳腺 磁共振成像 影像组学 |
| 英文关键词:carcinoma, ductal, breast magnetic resonance imaging radiomics |
| 基金项目: |
|
| 摘要点击次数: 367 |
| 全文下载次数: 89 |
| 中文摘要: |
| 目的 观察基于动态增强MRI(DCE-MRI)影像组学所示肿块样非特殊型浸润性乳腺癌(IBC-NST)单一形态学特征及其联合识别病灶导管原位癌(DCIS)成分的价值。方法 回顾性纳入67例肿块样IBC-NST,根据术后病理所见将28例共34个病灶纳入伴导管内癌成分DCISC组,39例共40个病灶纳入无DCISC组。基于乳腺DCE-MRI第2期轴位图像影像组学分析提取病灶14个形态学特征参数,比较组间病灶MRI特征及形态学特征;基于组间有统计学差异的形态学参数,以不同联合方式构建多因素logistic回归模型,利用受试者工作特征(ROC)曲线及其下面积(AUC)评估各单一参数及logistic回归模型识别肿块样IBC-NST伴DCIS成分的效能,并以DeLong检验加以比较。结果 DCISC组病灶形状不规则占比(64.71%)高于无DCISC组(25.00%,P＜0.05)。DCISC组病灶伸长率、平整度、最小轴长度、网格体积、球形度及体素体积均低于,而表面积体积比高于无DCISC组(P均＜0.05)。以单一IBC-NST病灶平整度及球形度识别DCIS成分的AUC(分别为0.812、0.793)均高于单一形状参数(AUC=0.659,Z=2.451、2.447,P均＜0.05)。以球形度+最小轴长度,平整度+最小轴长度+表面积体积比,以及前三者+伸长率各参数联合建立的logistic回归模型1、2、3识别DCIS成分的AUC分别为0.836、0.849及0.857。模型3的AUC与单一平整度、球形度差异无统计学意义(P均＞0.05)而高于其他单一形态学参数(P均＜0.05)。结论 基于肿块样IBC-NST单一DCE-MRI影像组学形态学特征及其联合可有效识别病灶伴DCIS成分,且效能均高于病灶MRI形状参数;联合模型诊断效能较部分单一特征模型有所提升。 |
| 英文摘要: |
| Objective To observe the value of morphological features of mass-like non-special type of invasive breast cancer(IBC-NST) based on dynamic contrast-enhanced MRI (DCE-MRI) radiomics alone and their combinations for identifying ductal carcinoma in situ (DCIS) components. Methods Sixty-seven patients with mass-like IBC-NST were retrospectively collected. According to postoperative pathological findings, 28 patients with 34 lesions were clustered into ductal carcinoma in situ component(DCISC) group, while the other 39 patients with 40 lesions were clustered into non-DCISC group. Fourteen parameters of morphological features within mass-like IBC-NST lesions were extracted from the 2nd phase axial images of breast DCE-MRI using radiomics analysis, and then MRI characteristics and morphological feature parameters were compared between groups. Multivariate logistic regression models were constructed under different combinations of morphological feature parameters being significantly different between groups. The performance of each parameter alone and logistic regression models for identifying mass-like IBC-NST with DCIS component were evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC), and DeLong tests were conducted for comparisons. Results The proportion of irregularly shaped mass-like IBC-NST in DCISC group (64.71%) was significantly higher than in non-DCISC group (25.00%, P＜0.05). In DCISC group, the elongation, flatness, least axis length, mesh volume, sphericity and voxel volume were all lower, while surface area to volume ratio was higher than those in non-DCISC group (all P＜0.05). AUC of flatness and sphericity for identifying mass-like IBC-NST with DCIS component (0.812 and 0.793, respectively) were both higher than that of lesion's shape (AUC=0.659, Z=2.451, 2.447, both P＜0.05). Logistic regression models (Model 1, 2 and 3) were established through combining sphericity + least axis length, flatness + least axis length + surface area to volume ratio, and the former three + elongation, with AUC of 0.836, 0.849 and 0.857, respectively. AUC of Model 3 was not statistically different with that of flatness nor sphericity (both P＞0.05) but higher than that of other morphological feature parameters alone (all P＜0.05). Conclusion Radiomics morphological features of mass-like IBC-NST on DCE-MRI alone or in combinations could effectively identify DCIS component, with performance superior to that of lesion’ MRI shape alone. Combining parameters model could improve diagnostic efficacy when compared to several norphological features alone. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
