A 4D Spatio-Temporal Model to Estimate Stroke Lesion Evolution on MR Perfusion-Diffusion Imaging Following Acute Ischemic Stroke

Islem Rekik¹, ², Stphanie Allassonnire², Stanley Durrleman³, Trevor Carpenter¹, Joanna Wardlaw¹

The 4D dynamic simulation of perfusion and diffusion lesion evolutions in acute-subacute ischemic stroke has considerable prognostic potential. We applied a current-based, 4D, dense, diffeomorphic regression model to perfusion and diffusion MR images acquired at three successive timepoints after stroke. The evaluation of the 4D patient-specific perfusion-diffusion evolution scenarios of 8 representative patients was promising as it fitted the reference standard manually-delineated lesion boundaries. Meanwhile, the dynamic spatio-temporal and kinetic contraction/expansion stroke lesion behavior did not fit with the expected lesion evolution as expected from the perfusion-diffusion mismatch hypothesis commonly used in stroke treatment.