Fig 1.
Cortico-cortical evoked potential analysis paradigms.
A: Convergent—Evoked responses at one chosen site (gray circle) are compared with the effect of stimulating all other sites (yellow circles with lightning bolt). For N electrodes, this characterizes N interactions. B: Divergent—The temporal response of all sites are examined and compared in response to stimulation of a chosen site (N interactions). C: All-to-all—All N2 interactions between sites are characterized. D: Hypothesis preselected—Two sites are chosen based upon a pre-defined anatomical or functional hypothesis, and a 1-way or 2-way interaction between them is characterized. E: In the convergent paradigm, all measured responses from a brain surface electrode are associated with the same underlying laminar architecture, so each response shape measured implies a distinct type of input. F: In the divergent paradigm, different shaped responses may be measured from different sites, in response to stimulation at a single site. This creates ambiguity because different shaped responses cannot distinguish between 1) the same type of output arriving at cortical sites with different underlying laminar architecture and 2) different types of inputs to sites with similar laminar architecture.
Fig 2.
Single-pulse cortico-cortical evoked potentials.
A: An array of brain surface (ECoG) electrodes were surgically placed on the left hemisphere of a brain tumor patient. B: The voltage (red trace) was measured at a parahippocampal gyrus (PHG) electrode site. C: Biphasic stimulation pulses were delivered between adjacent electrodes throughout the array (gray shows all stimulation pulse trials for stimulation at each site, red shows average). D: Responses from each stimulation pulse are aligned into a matrix Vk(t). E: Averaged subgroup responses Gn(t) (i.e. CCEPs, from the PHG measurement site) are shown between the two electrode sites that were stimulated to produce them.
Fig 3.
Technique for identification of basis profile curves (BPCs).
An illustration of the series of steps to extract BPCs from the voltage matrix V and subgroup assignments k ∈ n. A: Within stimulation-pair subgroup self-projections (all trials are projected into one another). B: Between-subgroup cross-projections. C: An illustration of sets of cross- and self-projections for stimulation-pair subgroup 11, S11,m. D: The significance of each set Sn,m is determined initially by t-value vs. zero. Negative t-values are set to zero. The matrix of these values is then scaled to 1, and labeled Ξ. E: Non-negative matrix factorization (NNMF) is performed to identify structure. F: The inner dimension of NNMF is iteratively reduced. G: BPCs are identified from the groups clustered in the rows of H.
Fig 4.
Projection of basis profile curves (BPCs).
A: The contribution of each BPC Bq to a single trial from its cluster can be quantified according to a scalar multiplier , and residual noise εk (trial 238 illustrated). B: The 3 BPCs for our example case. C: The spatial representation of BPCs, color-coded, with diameter and color intensity indicating magnitude (group-averaged signal-to-noise ratio). White circles show actual electrode locations and BPC projection magnitudes are placed at the the spatial average of the positions of the two stimulated electrodes. Each BPC distribution is individually scaled to maximum. Gray indicates sites discarded by thresholding (Fig 3G).