Dynamic contrast enhanced - MRI efficiency in detecting embolization-induced perfusion defects in a rabbit model of critical-limb-ischemia.

Critical limb ischemia (CLI) is a severe disease which affects about 2 million people in the US. Its prevalence is assessed at 800/100,000 population. However, no reliable tools are currently available to assess perfusion defects at the muscle tissue level. DCE-MRI is a technique that holds the potential to be effective in achieving this goal. However, preclinical studies performed with DCE-MRI have indicated low sensitivity assessing perfusion at resting state. To improve these previous results, in this work we propose new methodologies for data acquisition and analysis and we also revisit the biological model used for evaluation. Eleven rabbits underwent embolization of a lower limb. They were imaged at day 7 after embolization using DCE-MRI, performed on a 4.7 T small imaging device. Among them, n = 4 rabbits were used for MRI sequence optimization and n = 6 for data analysis after one exclusion. Normalized Areas under the curve (AUCn), and kinetic parameters such as Ktrans and Vd resulting from the Tofts-Kety modeling (KTM) were calculated on the embolized and contralateral limbs. Average and heterogeneity features, consisting on standard-deviation and quantiles, were calculated on muscle groups and whole limbs. The Wilcoxon and Fisher-tests were performed to compare embolized and contralateral regions of interests. The Wilcoxon test was also used to compare features of parametric maps. Quantiles of 5 and 95% in the contralateral side were used to define low and high outliers. A P-value <0.05 was considered statistically significant. Average features were inefficient to identify injured muscles, in agreement with the low sensitivity of the technique previously reported by the literature. However, these findings were dramatically improved by the use of additional heterogeneity features (97% of total accuracy for group muscles, P < 0.01 and 100% of total accuracy for the total limbs). The mapping analysis and automatic outlier detection quantification improvement was explained by the presence of local hyperemia that impair the average calculations. The analysis with KTM did not provide any additional information compared to AUCn. The DCE technique can be effective in detecting embolization-induced disorders of limb muscles in a CLI model when heterogeneity is taken into account in the data processing, even without vascular stimulation. The simultaneous presence of areas of ischemia and hyperemia appeared as a signature of the injured limbs. These areas seem to reflect the simultaneous presence of infarcted areas and viable peripheral areas, characterized by a vascular response that is visible in DCE.