Quantitative estimation of thrombus-erythrocytes using MRI. A phantom study with clot analogs and analysis by statistic regression models

Abstract

BackgroundThrombus composition has the potential to affect acute ischemic stroke (AIS) treatment.ObjectiveTo evaluate in an in vitro test the correlation of clot composition, especially erythrocytes (red blood cells (RBCs)), with the variation of signal intensity ratio (SIR) obtained with MRI sequences used for AIS, and qualification of the susceptibility vessel sign effect using clot analogs.Materials and methodsNine ovine clots were fixed in a gelatin-manganese solution and studied by MRI (T2GE, T2-weighted gradient echo; SWI, susceptibility-weighted imaging; FLAIR, fluid attenuated inversion recovery). RBC concentration was estimated using regression models (SLR, single linear regression; MLR, multiple linear regression; RF, random Forest; and ANN, artificial neural networking), which combined the SIR–histology relationship of three MRI sequences.ResultsNegative correlation was found between SIR and RBC concentration. T2GE SWI could not statistically distinguish clots with RBC content >54% and <23%. SLR was applied only to FLAIR images since T2GE and SWI demonstrated signal saturation. All four regression models showed a correlation between MRI and histology: SLR=0.981; MLR=0.986; RF=0.994, and ANN=0.971. One unknown clot was studied and agreement between SIR and histological analyses was found in all models.ConclusionsWe presented a method to quantify RBC concentration in clot analogs, combining SWI, T2GE, and FLAIR. This in vitro study has some limitations, so clot collection after thrombectomy with simultaneous imaging analysis is necessary to validate this model.