Multimodal lesion network mapping to predict sensorimotor behavior in stroke patients

Abstract

AbstractLesion network mapping (LNM) has proved to be a successful technique to map symptoms to brain networks after acquired brain injury. Beyond the characteristics of a lesion, such as its etiology, size or location, LNM has shown that common symptoms in patients after injury may reflect the effects of their lesions on the same circuits, thereby linking symptoms to specific brain networks. Here, we extend LNM to its multimodal form, using a combination of functional and structural connectivity maps drawn from data from 1000 healthy participants in the Human Connectome Project. We applied the multimodal LNM to a cohort of 54 stroke patients with the aim of predicting sensorimotor behavior, as assessed through a combination of motor and sensory tests. Test scores were predicted using a Canonical Correlation Analysis with multimodal brain maps as independent variables, and cross-validation strategies were employed to overcome overfitting. The results obtained led us to draw three conclusions. First, the multimodal analysis reveals how functional connectivity maps contribute more than structural connectivity maps in the optimal prediction of sensorimotor behavior. Second, the maximal association solution between the behavioral outcome and multimodal lesion connectivity maps suggests an equal contribution of sensory and motor coefficients, in contrast to the unimodal analyses where the sensory contribution dominates in both structural and functional maps. Finally, when looking at each modality individually, the performance of the structural connectivity maps strongly depends on whether sensorimotor performance was corrected for lesion size, thereby eliminating the effect of larger lesions that produce more severe sensorimotor dysfunction. By contrast, the maps of functional connectivity performed similarly irrespective of any correction for lesion size. Overall, these results support the extension of LNM to its multimodal form, highlighting the synergistic and additive nature of different types of imaging modalities, and the influence of their corresponding brain networks on behavioral performance after acquired brain injury.