Biomedical Engineering Reference
In-Depth Information
FIgURE 2.11:
Tuning depth for all the neurons computed from three kinematics respectively. The
upper plot is tuning depth computed from positions. The middle plot is for velocities. And the bottom
plot is for accelerations.
counting the number of times that each cell fires in a window. One can expect that for BMI decod-
ing, highly tuned cells will be more relevant. But tuning depth is not the only parameter of interest
for a BMI because the input must also supply sufficient information to reconstruct all of the kine-
matics in the desired trajectories, that is, the set of deeply tuned neural vectors must span the space
of the desired kinematic variable.
2.7.5 Information Theoretic Tuning Metrics
The tuning depth is a very coarse metric for the evaluation of the neuron receptive properties. The
mutual information between neural spikes and direction angles can alternatively embody the idea of
an information theoretic metric to evaluate if a neuron is tuned to a certain kinematic variable such
as the angle of the velocity vector.
p
(spike|angle)
∑
∑
I
(spike; angle) =
p
(
p
(angle)
(spike|angle) log
)
2
p
(spike)
(2.2)
angle
spike = 0,1
