Biomedical Engineering Reference
In-Depth Information
performance of linear models for BMIs improves as more neurons are recorded, but that the perfor-
mance improvement must be coupled with the
right
neurons. Irrelevant neurons can add to the model
bias and variance, thus reducing performance. One must also remember that if channel selection is used
the reduced number of cells makes BMIs more sensitive to the instability of neuronal firings over time.
Recent studies are showing that the activity of individual neurons and cortical areas used in BMI exper-
iments can vary considerably from day to day [
34
]; therefore, the variance over time of the importance
of neurons must be quantified in future studies. It is therefore not clear how practical the neuron selec-
tion techniques will be in the surgery stage. For these reasons, we advocate the use of a higher sampling
of the cortical activity to help improve this ratio until other models are proposed that take advantage of
the information potentially contained in the spike trains and not exploited by linear models.
In this model-based framework, it is clear that better BMIs can be built by combining regu-
larization techniques with a subset of important cells. This question is rooted in our goal to build
models for BMIs that generalize optimally. It should not be forgotten that regularization and chan-
nel selection is biased by the type of model chosen, its performance level, and by noise in the data.
Therefore, it is important to pursue a model-independent approach to preprocess the data. We hy-
pothesized from a combined neurophysiological and modeling point of view that highly modulated
neurons spanning the space of the kinematic parameter of interest should be chosen. Intuitively,
these constraints make sense for the following reasons: If a cell is modulated to the kinematic pa-
rameter, the adaptive filter will be able to correlate this activity with the behavior through training.
Otherwise, neural firing works as a broadband excitation that is not necessarily related to better
performance. If a group of cells are highly modulated to only a part of the space, the adaptive filter
may not be able to reconstruct data points in other parts of the space.
A final comment goes to the overall performance of the BMI system built from adaptive models.
Although it is impressive that an optimal linear or nonlinear system is able to identify the complex
relations bnetween spike trains in the motor cortex and hand movements/gripping force, a correlation
coefficient of ~0.8 may not be sufficient for real-world applications of BMIs. Therefore, further research
is necessary to understand what is limiting the performance of this class of adaptive linear and nonlinear
systems. Another issue relates to the unrealistic assumption of stationarity in neural firings over record-
ing sessions that is used to derive results presented in this. In future studies, it will be necessary to assess
the time variability of the neuronal rankings and determine its effect on model generalization.
REFERENCES
1. GemanS., E. Bienenstock, E., and R. Doursat,
Neural networks and the bias/variance dilemma
.
Neural Computation, 1992.
4
: pp. 1-58.
2. Sanchez, J.C., et al.,
Input-output mapping performance of linear and nonlinear models for estimating
hand trajectories from cortical neuronal firing patterns
, in International Work on Neural Networks
for Signal Processing. 2002. IEEE, Martigny, Switzerland.
doi:10.1109/NNSP.2002.1030025
