Human cognitive neuroscience seminar

Speaker: Ding Junhua (University of Edinburgh)

Title: A multiverse evaluation of connectome-based lesion-symptom mapping in post-stroke aphasia

Abstract: Connectome-based lesion-symptom mapping (CLSM) relates brain connectivity disruption to behavioural impairments. CLSM can be based on different image modalities (DWI vs T1), regions of interest (whole brain vs language network), and measure types (tracts, parcel-based connectome, or graph theory). We used multiverse analysis to compare these processing choices and identify factors that influence the results. The analyses used T1 and DWI images and Boston Naming Test performance from 50 participants with aphasia following left hemisphere stroke. “Direct” measures were derived from DWI using probabilistic fibre tracking. “Indirect” measures were derived by overlaying lesion masks (from T1) on a white matter template. They were further constructed for the whole brain (all AAL parcels) and the language network (14 LH AAL parcels). Numerous tract-based and graph theory metrics were calculated. In sum, the image modality, regions of interest and metric type all led to a poor convergent validity. Metrics showed weak correlations and inconsistent CLSM results, despite nominally measuring the same construct. Substantial methodological work is needed to validate CLSM metrics. In the interim, pre-registration and multiverse analysis are useful strategies for addressing the flexibility and inconsistency across different approaches to CLSM.


The seminars are organised by the Human Cognitive Neuroscience research group. For further information, or if you would like to join the e-mail list for these seminars, please email Ed Silson.

Ed Silson

Human cognitive neuroscience

Mar 09 2022 -

Human cognitive neuroscience seminar

2022-03-09: A multiverse evaluation of connectome-based lesion-symptom mapping in post-stroke aphasia

Online via link invitation