动态对比增强不同时间分辨率下MRI线性参照模型和双室Tofts模型鉴别乳腺良、恶性病变的价值

Linear reference region model and Tofts model in dynamic contrast-enhanced MRI of discriminating benign and malignant breast lesions comparative study

目的探讨动态对比增强（DCE-MRI）不同时间分辨率下线性参照模型（LRRM）和双室Tofts模型鉴别乳腺良、恶性病变的价值。方法回顾性分析诊断经穿刺活检或手术病理证实，行DCE—MRI检查的85例女性乳腺肿块型患者，其中良性病变37例、乳腺癌48例。患者均行双侧乳腺DCE—MRI检查，其中31例（恶性病变15例、良性病变16例）采用低时间分辨率（扫描时间18s／期），54例（恶性病变33例、良性病变21例）采用高时间分辨率（扫描时间7s／期）。将患者病变分为恶性组、良性组和正常组（良性病变组对侧正常腺体），测量不同时间分辨率下乳腺病变的容量转移常数（Ktrans）。采用Kruskal—WallisH秩和检验比较正常组、良性组和恶性组间Ktrans值的差异，并以病理结果为金标准，绘制不同时间分辨率条件下2种模型Ktrans-值鉴别诊断乳腺良、恶性病变的ROC曲线，评价其诊断效能。结果高、低时间分辨率下，正常组Tofts模型的Ktrans值分别为（0．048±0．022）、（0．090±0．040）／rain，LRRM的Ktrans值分别为（0．301±0．197）、（0．287±0．225）／min；良性组Toffs模型的Ktrans值分别为（0．289±0．163）、（0．211±0．080）／min，LRRM的Ktrans值分别为（0．624±0．358）、（0．593±0．165）／min；恶性组Toffs模型的Ktrans值分别为（0．959±0．451）、（0．524±0．285）／min，LRRM的Ktrans值分别为（1．576±0．935）、（0．956±0．180）／min。3组间不同时间分辨率下的Ktrans值差异均有统计学意义（P均〈0．05）。高时间分辨率Tofts模型、高时间分辨率LRRM、低时间分辨率Toffs模型、低时间分辨率LRRM的Ktrans值鉴别乳腺良、恶性病变的ROC曲线下面积分别为0．941、0．876、0．850、0．933，分别以Ktrans值为0．304、0．917、0．252、0．789／min为诊断界值，鉴别乳腺良、恶性病变的敏感度和特异度分别为93．9％、85．7％，80．0％、90．5％，80．0％、81．2％，80．0％、87．5％。结论低时间分辨率下LRRM较Torts模型的定量参数Ktrans对诊断乳腺癌价值更大，高时间分辨率下Toffs模型Ktrans值诊断乳腺癌的敏感度、特异度优于LRRM模型。

Objective To investigate and compare the diagnostic values of linear reference region model （LRRM） and Extended Torts model in quantitative dynamic contrast-enhanced magnetic resonance imaging （DCE-MRI） analysis of benign and malignant breast lesions under different temporal resolutions. Methods Eight five suspicious breast cancer women underwent bilateral DCE-MRI exam, 37 patients were benigns and 48 patients were malignants. Among those, 31 patients（15 malignant, 16 benign） were scanned with 18 s per phase, and 54 patients（33 malignant, 21 benign） were scanned with 7 s per phase, and they were assigned into breast cancer group, benign lesion group and healthy gland group proven by surgery orbiopsies. For the same model,Ktrans values of the three groups under different time resolution were first analyzed using Kruskal-Wallis H rank sum test. Receiver operator curve （ROC） was used to analyse the diagnostic efficiency of Ktrans values. Results Under high and low temporal resolutions, Ktrans values of the healthy group were （0.048 ±0.022） and （0.090 ±0.040）/min for extended Tofts model,（0.301 ± 0.197） and （0.287±0.225）/min for LRRM model respectively.Ktransvalues of the benign group were （0.289±0.163） and （0.211 ± 0.080）/min for extended Tofts,（0.624± 0.358） and （0.593 ± 0.165）/min for LRRM respectively.Ktrans values of the malignant group were （0.959±0.451） and （0.524±0.285）/min for extended Tofts,（1.576±0.935） and （0.956±0.180）/min for LRRM respectively.There were significant differences among the three different groups（P〈0.05）.Area under the ROC to differentiate benign and malignant breast lesions for Extended Toffs in high temporal, LRRM in high temporal, Tofts in low temporal and LRRM in low temporal were 0.941, 0.876, 0.850 and 0.933, with Ktrans cutoff values of 0.304, 0.917, 0.252 and 0.789/min,and sensitivity of 93.9%, 80.0%,80.0%, 80.0%;speeificity of 85.7%, 90.5%, 81.2%, 87.5% respeetively. Conclusion Under low temporal resolutions, Ktransof LRRM model had better sensitivities and specificities in differentiation of benign and malignant breast lesions than Extended Toffs model, which was the opposite in high temporal resolutions.