Processing for quantitative parameters in rodent pancreatic MRI
MRI quantitative parameters can be obtained from advanced image processing methods using machine learning algorithms[35-37]. For practical considerations, quantitative parameters are re-generated from in-house built Matlab programs using non-linear least square methods with CPU (Central Processing Unit) acceleration.
T2 mapping: Traditionally, the transverse relaxation time T2 is obtained using multi-echo spin-echo pulse sequences, by sampling signals at several different echo-times (TE), and fitted to either multi- or single-exponential decay functions. Fast T2 mapping can be obtained using balanced steady-state free precession (SSFP) readout.
T1 mapping: On a clinical scanner, fast T1 mapping can be measured using inversion recovery methods or from variable flip angles experiments[39,40]. Since the MRI acquisition has to be synchronized with animal respiration, the effective repetition time (TR) is usually longer than 1 s. Thus, inversion recovery based protocols would be suggested for T1 mapping in rodent pancreatic imaging. Typically, the equation for measured signal in the inversion recovery T1 mapping experiment is a three-parameter function: SI(t) = a + b × exp(-t/T1*), where SI(t) is the signal intensity after each inversion time t, and T1* is the effective longitudinal relaxation time. The actual T1 relaxation time can be obtained after correction for the flip angles, or the Look-Locker readout.
Diffusion-weighted model: In DWI experiments, the simple Gaussian diffusion can be assumed using a mono-exponential model. The two-compartment intravoxel incoherent motion model on the other hand is currently widely used in clinical pancreatitis and pancreatic tumor studies[43,44], and separates diffusion into the true-diffusion and the pseudo-diffusion fraction. Alternatively, sampling with high b-values above 1000 s/mm2 can be applied for non-Gaussian diffusion estimation using a diffusion kurtosis model.
Post-processing for DCE model: The first step in DCE data post-processing is the conversion of the raw MRI signal to the tissue concentration time curve (CTC). The tissue concentration Ct of contrast agent (CA) during the DCE perfusion experiment is solved as: 1/T1(t) = 1/T10 + r1 × Ct(t), where T10 is the T1 value before contrast injection, obtained from inversion recovery T1 mapping, and r1 is the longitudinal relativity of the applied CA. In a high temporal resolution DCE experiment, the T1 relaxation T1(t) after CA injection can be converted from the signal intensity (SI) time curve as described previously. Alternatively, CTC information can be directly extracted from the dynamic T1 mapping.
The vascular input function (VIF) Cp is determined by the CA concentration in blood Cb: Cp = Cb/(1 - Hct), which is obtained from CTC of the aorta or a major vein, and the hematocrit level Hct which is set to 42% in our studies. VIF is usually fitted into a bi-exponential function for further kinetic modeling. Perfusion indices, the transfer coefficient Ktrans and the rate constant kep, can be obtained from the standard or the modified Tofts model. In practice, the discrete convolution can be constructed as a matrix multiplication. The fraction volume ve of extravascular extracellular space is calculated as: ve = Ktrans/kep.