Prediction of tissue outcome after experimental stroke using MRI-based algorithms

Abstract

Acute ischemic stroke is a major cause of death and disability in modern western society. Possible benefit of the only clinically approved therapeutic intervention, i.e. thrombolysis, is complicated by complex pathophysiologic interplay of factors resulting from the ischemic insult. As a result, strict therapeutic guidelines limit the applicability of this therapy. Yet these guidelines may not always apply on an individual patient level, thereby unnecessarily excluding patients who could have benefitted from therapeutic intervention. Therefore early and accurate assessment of tissue injury and the prediction of its progression are crucial for individualized strategic therapeutic planning. In this thesis ‘Prediction of tissue outcome after experimental stroke using MRI-based algorithms’ the potentials of novel MRI-based models for prediction of brain tissue outcome after ischemic stroke have been evaluated. Serial MRI data, acquired in different experimental animal models of stroke, were employed to evaluate the potential of various multiparametric statistical models for monitoring of tissue injury progression and prediction of tissue outcome. The studies in this thesis demonstrate that the use of novel voxel-based prediction methods provide significantly improved insights in brain tissue injury progression, with the specific ability to predict variable degrees of cerebral damage. Our data revealed that: (1) multiparametric clustering techniques allow for improved differentiation of tissue injury progression as compared to previously employed volumetric methods; (2) benefit from reperfusion was more specifically predicted by an angiography-diffusion mismatch than a perfusion-diffusion mismatch; (3) specific tissue outcome prediction algorithms enabled infarction risk-based differentiation of tissue amenable for reperfusion from irreversibly injured tissue; and (4) ischemic areas with subsequent hemorrhage were more accurately predicted by multiparametric prediction models than single MRI-parameter thresholding-based methods. The findings in this thesis show that voxel-wise integration of a multitude of (MR imaging-based) biomarkers, representing complex and heterogeneous disease mechanisms within a single easily interpretable index, give the ability to differentiate, predict, and track heterogeneous tissue progression without the need of defining restricted viability thresholds. This offers exciting prospects for multiparametric algorithms in (pre-)clinical stroke treatment trials and points toward appealing opportunities for improved personalized health-care in the near future.