Statistical evaluation of fitting models of diffusion tensor imaging in characterizing normal porcine myocardium

It has been long believed that the decay of diffusion-attenuated signal in MRI complied with monoexponential manner. However, a number of researches found that diffusion signal tended to deviate pronouncedly from monoexponential model with increase of b values in some tissues, such as brain, kidney and prostate. The purpose of the current study was by fitting in each diffusion direction separately in diffusion tensor imaging (DTI), to analyze whether a mono- or biexponential or diffusion kurtosis imaging model gives the best fitting model of DTI in describing diffusion characteristics in normal porcine myocardium. Five non-zero b values between 500 and 2000 s/mm² were selected and their corresponding myocardial diffusion signal used to fit the three diffusion models in each diffusion direction separately. Statistical tests showed that biexponential model fitted better than the diffusion kurtosis imaging model, and monoexponential model gave the worst fitting performance. The results suggest that diffusion tensor imaging of myocardium may be modeled biexponentially to derive accurate characterization of myocardial microstructure.