Dear all,
We want to apply DCM analysis in a post-hoc manner to investigate how a big brain network works, the maximal number of nodes is one of the major concerns. I noticed that in a recent paper “Network discovery with large DCMs”, a large network with 20 nodes was investigated using post-hoc DCM. But if the B and C matrix is also specified, such a large network is hard to achieve.
The full model of deterministic DCM in our study is specified as follow. The intrinsic connection matrix include all the connections including those to themselves. There are three inputs, one is the sensory input that only acts on the primary sensory nodes, the other two are modulatory inputs that only act on cross-node connections. Please see the attachment for an example of the specified A,B and C matrix. However, it seems that the maximal number of nodes in such a DCM is about 12~13, and the “spm_nlsi_GN” routine would report convergence failure if the number of nodes is larger than this in our situation. We want to increase the size of network up to ~18 nodes to include all the regions we are interested in. Is it possible to achieve this goal? Any advice and suggestions will be greatly appreciated.
By the way, the full timecourse consists of 560 data points with a TR of 2s. The VOIs were extracted using SPM with a threshold of voxel-level uncorrected p < 0.05 and cluster-size k > 20.
Best regards,
Bingjiang
|