creating the problem statement

[jessb0t]

conservation of energy

increase in high frequency = decrease in low, but increase in low frequency in surrounding regions

amplification of motor signal

How does sensory feedback (visual, somatosensory) mediate the relationship between task instruction and signal in ctx?

model performance imagined > real

How well does a model optimized for imagined movement perform on real movement?

similarity analysis between imaginary and real signal

Is it possible to extract the imaginary signal out of the real, overt signal?

speed of motor skill learning

Is there higher activation of electrodes over the duration of the imagined session? Higher synchrony of hand/mouth representations across multiple electrodes? How does visual feedback alter the synchrony?

[jessb0t]

diagram of amplification of motor signal:

[jessb0t]

José

Think critically about measurement/benchmarks: how will we define 'good,' 'accurate,' or 'similar?' The relevant literature can help inform this choice, and it must be done before we start running any models.

Jason

• Nyquist Frequency: Given our sampling rate of 1000 Hz, we cannot look for effects that would be a frequencies higher than 500 Hz. As an example, the timing of an action potential is ~1 ms (=1000 Hz) and therefore is beyond the view of our data. However, activities over a 10 ms (100 Hz) or 100 ms (10 Hz) time course are accessible to us.
• we could consider pre-training a BCI decoder on several subjects, then testing on a held-out subject to tease apart subject-specific versus general patterns
• we could look into temporal sequencing: do we see that the imagined portion of the action is actually part of the pre-cursor (that is, it is the same and then the overt movement begins)? -- this would give us a "signature of motor planning as distinct from motor execution"
• think about our project at three levels of hypothesis: the scientific hypothesis (very general, just about brains, not even at the ECoG level), the experimental hypothesis (essentially what we have been doing thus far), and the statistical hypothesis (the numerical model)
• avoid "yes/no" questions and use "how much" questions as these are ultimately more interesting (and help you define your statistical hypothesis)

[kurtlehn1]

Another interesting thing Jason mentioned which relates to the levels of hypotheses was to think mechanistically: e.g. there is literature that suggests that high frequency gamma activity relates to neuronal spiking and low frequency activity relates neurotransmitter effects which act over longer time scales to change local field potentials, so perhaps there is some change in low frequency acitivity (i.e., local field potential) that precedes high frequency gamma activity (i.e., spiking) that is a signature of motor planning.

thinking more about it, this signal would likely be different in real vs. imagined movements.

[jessb0t]