Reconstructing time-domain data from discontinuous Percept™ PC and RC output using external data acquisition and linear filtering

Abstract

Background

The Medtronic Percept™ PC and RC are deep brain stimulation (DBS) systems with recording capability. However, when the stimulation frequency is changed, the recordings were segmented, introducing interruptions that shift each segment in the time domain.

New method

Ex-vivo, stimulation frequency was changed while local field potential was being recorded in both leads. One lead captured stimulation artifacts from the DBS system, and another captured stimulation artifacts from the DBS system plus a 0.5 Hz impulse train from an external stimulator. Timing errors were assessed by comparing Percept™-recorded impulses to the gold-standard external stimulator impulses. The Percept™ recordings were then time-shifted to match the external system’s timing, based on the time difference between the two systems when stimulation frequency change was indicated.

Results

For both PC and RC, the sawtooth pattern occurred. Timing errors were noted to have linear ramps interrupted by sudden drops, which were used to develop an algorithm to correct, leveraging occasions where the Percept™ happens to record the true moments of stimulated frequency change. Errors ranged from -400 to 400 ms for PC, and from -1 to 1 s for RC. The timing reconstruction algorithm reduced the error to -10.07 ± 45.06 ms (mean ± std) for PC, and -23.52 ± 17.32 ms (mean ± std) for RC.

Comparison with existing methods

We measure and characterize the timing errors of each recorded segment, using ex-vivo DBS hardware, and propose a strategy to correct them.

Conclusion

This approach can be applied in-vivo using electroencephalogram to correct timing errors that are significant with long recordings, enabling accurate time synchronization.