CAJ | 학술논문

目的对比双指数弥散模型和弥散峰度模型对正常前列腺组织的弥散加权成像在1.5 T不同b值分段下的拟合效果。材料与方法采集11例健康男性前列腺T2加权成像和弥散加权磁共振的多b值成像，从0到3000 s/mm2中选取31个b值，将两个弥散模型对弥散加权信号原始数据作曲线拟合，根据曲线拟合计算的拟合信号强度值与原始图像采集的信号值的差值大小，选取对某个模型该差值都较小的对应的连续多个b值，组成一个分段，该模型则为此b值分段最优的拟合模型。结果在所有b值数据的整体拟合中，双指数弥散模型调整后的决定系数R2大于弥散峰度模型。b值在500～1000 s/mm2时，弥散峰度模型的均方根误差比双指数弥散模型更小；b值在0～500 s/mm2和1000～3000 s/mm2时，双指数弥散模型的均方根误差比弥散峰度模型更小。结论双指数弥散模型和弥散峰度模型对正常前列腺的弥散加权成像在不同的b值下的拟合优度不同，不同b值分段具有不同的最优拟合模型，将两个模型结合起来对前列腺弥散加权成像进行分析可能会为临床的诊断提供更多的帮助。
목적대비쌍지수미산모형화미산봉도모형대정상전렬선조직적미산가권성상재1.5 T불동b치분단하적의합효과。재료여방법채집11례건강남성전렬선T2가권성상화미산가권자공진적다b치성상，종0도3000 s/mm2중선취31개b치，장량개미산모형대미산가권신호원시수거작곡선의합，근거곡선의합계산적의합신호강도치여원시도상채집적신호치적차치대소，선취대모개모형해차치도교소적대응적련속다개b치，조성일개분단，해모형칙위차b치분단최우적의합모형。결과재소유b치수거적정체의합중，쌍지수미산모형조정후적결정계수R2대우미산봉도모형。b치재500～1000 s/mm2시，미산봉도모형적균방근오차비쌍지수미산모형경소；b치재0～500 s/mm2화1000～3000 s/mm2시，쌍지수미산모형적균방근오차비미산봉도모형경소。결론쌍지수미산모형화미산봉도모형대정상전렬선적미산가권성상재불동적b치하적의합우도불동，불동b치분단구유불동적최우의합모형，장량개모형결합기래대전렬선미산가권성상진행분석가능회위림상적진단제공경다적방조。
Objective:To compare the fitting behavior of biexponential diffusion model and diffusion kurtosis model (DKI) on diffusion-weighted imaging (DWI) of healthy prostate at different b-values at 1.5 T.Materials and Methods:T2-weighted imaging and DWI of prostate was performed on 11 healthy man with 31 b-values ranging from 0 to 3000 s/mm2. The DWI signals were iftted into two diffusion models. The continuous b-values with smaller deviation between the signal intensities of iftted curves of one model and the acquired original data can be combined into one section. The best iftting model could be selected for the b-values sections.Results:The adjusted R2 of the full b-values for biexponential diffusion model was bigger than that for DKI. The RMSE of DKI was smaller than biexponential model with b-values from 500 to 1000 s/mm2, and the RMSE of biexponential diffusion model was smaller than DKI with b-values ranging from 0 to 500 s/mm2 and from 1000 to 3000 s/mm2.Conclusion:For DWI of prostate, biexponential diffusion model and DKI behaviors diversely on the goodness of iftting at different b-values. The different b-values sections can be iftted best with different diffusion models. It may potentially provide more help for clinical diagnosis when combining the biexponential diffusion model and DKI.