Abstract:

This paper investigated the impact of different b-value acquisition ranges (from 0 to 2 500 s/mm²) on the quantitative parameter calculation of six body diffusion models, including mono-exponential (Mono), diffusion kurtosis imaging (DKI), intravoxel incoherent motion (IVIM), stretch exponential model (SEM), fractional-order calculus model (FROC), and continuous time random walk model (CTRW). The influence of different b-value acquisition ranges on parameter calculation was evaluated through correlation between diffusion model parameters, t-test, and the ability to differentiate benign and malignant prostate lesions. The results showed that compared with that of the reference sampling scheme (0~2 500 s/mm²), the difference of the mean value of region of interest (ROI) with the same diffusion parameters gradually increases as the maximum b-value decreases, but the correlation decreases only slightly, and the ability to differentiate between benign and malignant prostate lesions remains at a similar level. Based on the experimental results, a b-value range of 0~1 500 s/mm² is recommended for clinical practice, because this scheme takes collection efficiency into account, the correlation of more than half of its parameters with those of the reference sampling scheme is not less than 0.98, and the difference in the values of area under the curve (AUC) between benign and malignant differentiation is less than 0.01. In addition, the sensitivity of different diffusion models to the b-value scheme varies, with the parameters of SEM and CTRW models being relatively less affected by the b-value range.

Key words: MR imaging, Data sampling, Diffusion models, FROC, CTRW