I would appreciate your advice on DCM model specification for a mixed block
and event-related fMRI study. The blocks are defined by two alternative
attentional states, ATT1 and ATT2. Within each atentional state subjects
perform several short trials. For each trial, a precue is presented, PRE,
2s before a probe cue, PROBE1 or PROBE2 and subjects make an appropriate
response RES(usually!).
The precue could be regarded as an extrinsic input, upstream in a
heirarchical network. The extrinsic effects of precues would essentially be
the same in all trials within a block. However, the effects of precuing on
the downstream connectivity within a network may vary according to
attention, which could presumably be modelled by a bilinear vector, ATT1 vs
ATT2. So far so good.
However, the two types of probe cue are necessarily randomly ordered within
a block, and I am not sure how best to model them.
1. should I model all probe events as a single extrinsic event type, PROBE,
but model the diffreential effects of PROBE1 vs PROBE2 as a second (event-
related) bilinear vector? If so, the PROBE and PRE cues have extrinsic
vectors that are very closely correlated, offset by just 2 s
2. Should I model all trials (PRE-PROBE-RES) within blocks as a single
extrinsic input to the system (dsespite the 2 delay) but now have two
biliear vectors, one describing the context/block effects of attention,
ATT1 vs ATT2, and the second describiing the event-related difference
between trials with PROBE1 vs trials with PROBE2?
3. Should I model all cue types as a single extrinsic stimulus
input 'STIM', with separate bilinear vectors to decxribe the interactions
between context and precue types (PRE-ATT1 vs PRE-ATT2) and the interaction
between PROBE and context (PROBE-ATT1 vs PROBE-ATT2) and probe type (PROBE1
vs PROBE2), perhaps with a higher order interaction bilinear vaector
(PROBE1-ATT1 PROBE1-ATT2 PROBE2-ATT1 PROBE2-ATT2) ?
or should one avoid paradigms with compound trials under different contexts?
thank you in advance,
James Rowe
|