Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Mapping the vascular territories in left hemisphere stroke
There is a Poster PDF for this presentation, but you must be a current member or registered to attend SNL 2023 to view it. Please go to your Account Home page to register.
Poster C55 in Poster Session C, Wednesday, October 25, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
Also presenting in Lightning Talks C, Wednesday, October 25, 10:00 - 10:15 am CEST, Auditorium
Ajay Halai1, Ying Zhao2, Cathy Price3, Matthew Lambon Ralph1; 1MRC Cognition and Brain Sciences Unit, University of Cambridge, UK, 2Department of Psychiatry, University of Cambridge, Cambridge, UK, 3Wellcome Centre for Human Neuroimaging, University College London, London, UK
Lesion-symptom mapping (LSM) has been a cornerstone of cognitive neuroscience since the 19th century, where Broca’s and Wernicke’s aphasia were first described. Classical neurology studies were limited in their ability to map precisely, in contrast MRI allows us to map with millimetre specificity. One major caveat that is rarely addressed relates to the fact that brain damage after stroke is not random but constrained by neurovasculature. It is known that LSM can be prone to mis-localisation (e.g., Mah et al., 2014), where the error is skewed towards the areas most likely to be damage (for middle cerebral artery (MCA) stroke, that is commonly the medial insula and subcortical regions). The problem is not alleviated with novel multivariate mapping methods (Sperber et al., 2019). This suggests an alternative approach is needed, however, before one can develop such an approach we need to be able to map the vascular territories. In a proof of concept study, we showed a coherent statistical structure in the lesion profiles of left hemisphere stroke aphasic patients (Zhao, Halai, et al., 2020). Importantly, the underlying structure mapped onto post mortem angiography studies of the MCA. The sample was, however, relatively small and we were unable to measure reliability. In this study, we used 526 left hemisphere stroke patients from the PLORAS database and applied hierarchical clustering analysis to the lesions. We chose this method because the neurovascular system is known to follow a hierarchical framework across three major arteries (anterior [ACA], MCA, posterior [PCA]). We first used minimum description length to determine the number of likely clusters (iterated 100 times across sub-samples). We estimated 28 clusters, which were then extracted as 3-D volumes. From a macro scale, we identified four large clusters ordered in the proximity from anterior to posterior: ACA, superior MCA, inferior MCA and PCA. This fits with known anatomy and expected nature of lesions in a typical stroke population. We used the areas described in the post mortem studies as validation to link to the regions identified in the current study – there was an extraordinarily high degree of conformity. We repeated the analysis for reliability (split-half), where we applied the same clustering algorithm to two random sub-samples. We matched clusters based on spatial overlap using DICE. In order to determine stability, we iterated this 1000 times with a random split-half each time. The average DICE for all clusters was 0.5 (SD=0.15; range=0.21-0.77). All but four clusters had DICE values above 0.4. DICE was correlated with the frequency of lesion to a cluster (r=0.69), suggesting that less stable clusters might be driven by fewer data points. We also found that the distance between clusters across split halves was stable and correlated (r=0.7), which means the overall shape and structure remained stable. We demonstrated the reliability and stability of a neurovascular atlas. This raises important questions for LSM in our field: e.g., 1) can we use knowledge of the statistical structure in correcting mis-localisation in LSM?; and 2) are these clusters our ‘effective’ resolution?
Topic Areas: Disorders: Acquired, Methods