guest ::
| Home > Publications database > Influence of different spike sorting algorithms on the detection of single unit spiking activity in thesubthalamic nucleus of patients with Parkinson’s disease |
| Home > Publications database > Influence of different spike sorting algorithms on the detection of single unit spiking activity in thesubthalamic nucleus of patients with Parkinson’s disease |
| Poster (Other) | FZJ-2016-05175 |
; ; ; ; ; ; ; ;
2016
Abstract: Objective: To determine the most suitable spike sorting algorithm for intraoperatively recordedextracellular currents in the subthalamic nucleus (STN) of patients with Parkinson’s disease (PD).Background: In PD the STN may play an important role in the generation of pathological oscillatoryactivity within the basal ganglia-cortex loop. Extracellular currents recorded from the STN of PDpatients, obtained during Deep Brain Stimulation surgery can provide important information aboutpathological spiking activity of STN neurons. A challenge in analyzing neural spike data is to assigndetected spike to individual neurons that contribute to the extracellularly recorded activity. To thisend, a group of algorithms commonly referred to as "spike sorting" is used. Today, numerous spikesorting algorithms are available, partly side-by-side in commercial spike sorting packages. However, itis unclear to which degree different algorithms yield similar sorting results.Methods: Spiking activity was recorded intraoperatively at multiple sites within the STN-area of 3awake PD patients. The recorded time series were sorted ("Plexon Offline-Sorter") by 2 semi-automatic (template-based (TB) and K-Means) and by 4 automatic algorithms (K-Means, Valley-Seeking (VS), standard- and t-distribution Expectation-Maximization). We took a multiple validationapproach by comparing the spike times in the detected single units (SU) to determine how thesorting procedure influences the subsequent spike train analysis.Results: Preliminary results demonstrated a wide variability in the number of detected SU betweenthe 6 sorting algorithms (26-62 SU in 21 channels). Additionally, the analysis of the spike trainsstatistics revealed a large variability in the inter-spike interval (between 52 +/-38 and 321 +/- 778(mean +/-SD in ms)) and the coefficient of variation (between 0.61 and 18.91). In summary, thoughnot optimal, TB- and VS-algorithms proved to be the most conservative among the sorting methodsinvestigated.Conclusion: Our results strongly argue for the need of a standardized validation procedure for spikesorting algorithms based on ground-truth data. Moreover, to ensure reproducibility of results andenable scientists to understand differences between results obtained from different experiments, adetailed description of spike sorting procedure becomes a necessity.
|
The record appears in these collections: |