Tuesday, 15 December 2020
Tuesday, 15 December 2020
In a letter to the editor, Dr Aaron Boes (2020) commented on our study (Salvalaggio et al., 2020) in which we com- pared different structural and functional MRI methods to predict behavioural deficits in a large cohort of subacute stroke patients (Corbetta et al., 2015). Specifically, we used lesion symptom mapping, i.e. the behavioural prediction based on lesion location and volume, as a baseline for three methods measuring brain network(s) disconnection. Two methods assessed disconnection ‘indirectly’ using the lesion to generate maps of altered connectivity based on healthy control datasets (7 T ‘Human Connectome Project’ datasets: http://www.humanconnectome.org/study/hcp-yo ung-adult/; Vu et al., 2015). We computed structural diconnection of white matter pathways (SDC) and function- al disconnection of brain networks (FDC), or lesion network mapping (Boes et al., 2015). The third method measured patterns of altered functional connectivity directly based on the temporal correlation of the spontaneous blood oxygen level-dependent (BOLD) signal. Our results showed comparably high behavioural prediction for lesion, SDC, and functional connectivity, but a weak prediction for FDC. We concluded that FDC might be used to localize abnormal networks, but its low sensitivity to the severity of behavioural deficits implies that it cannot predict behaviour or recovery of function, nor be a substitute for direct functional MRI functional connectivity...
Whole brain mean functional connectivity strength (Pearson’s correlation) for each tissue regions of interest shows a different distribution pattern: stronger values (e.g. 40.20) are more frequent in the grey matter seed map compared to the white matter seed map (distributions of 0 values are truncated to improve visualization).
Reply: Lesion network mapping: where do we go from here?