Session Time: 1:45pm-3:15pm
Location: Hall 3FG
Objective: To detect the early-onset of high-voltage spindles (HVSs), with a minimum latency to realize smart neurostimulators for treating Parkinson’s disease (PD)
Background: Resting tremor (4-7 Hz) is the most prevalent motor symptom in at least 69-100% of PD patients. It is causally related to the abnormally-synchronized, HVSs in the basal ganglia-thalamocortical network. HVSs are rhythmic, spike-and-wave, oscillations at 5-12 Hz in the local field potentials (LFPs). Deep brain stimulation (DBS) is a well-recognized treatment for PD. However, the continuous electrical stimulation in contemporary DBS systems induces adverse neuropsychiatric side-effects. Therefore, “closed-loop” DBS is preferred to deliver stimulation only on the detection of pathological neural oscillations.
Methods: LFPs are recorded from eight brain regions of five 6-hydroxydopamine-lesioned PD rats. A low-order, frequency-selective, autoregressive model at interval is formulated from the LFPs, and its parameters are trained by the Kalman filter offline. Given a very-short raw LFP segment (144 ms), the trained model predicts its temporal and spectral features, which are classified by HVS detector using 16-point fast Fourier Transform (FFT).
Results: The algorithm is evaluated with 2384 LFP segments recorded from different brain regions and different PD rats. It achieves a satisfactory performance (84.3% sensitivity, 87.3% specificity, 85.8% accuracy) with latency less than 150 ms. The low-dimensional predictive model and simple detection mechanism facilitates hardware-friendly implementation in low-power embedded systems, such as, 8-bit microcontrollers.
Conclusions: The simulation results suggest that the proposed algorithm can be used to design smart neurostimulators for realizing closed-loop DBS systems to mitigate the pathological effects of HVSs, such as the resting tremor. The training method can also be tuned to learn desired pathological features relevant to PD (beta activity) or other neural disorders such as schizophrenia, Alzheimer’s disease, and epilepsy.
To cite this abstract in AMA style:R. Perumal, V. Vigneron, C. Chuang, Y. Chang, S. Yeh, H. Chen. An efficient algorithm for predicting pathological high-voltage spindles related to Parkinsonian resting tremor from local field potential recordings [abstract]. Mov Disord. 2018; 33 (suppl 2). https://www.mdsabstracts.org/abstract/an-efficient-algorithm-for-predicting-pathological-high-voltage-spindles-related-to-parkinsonian-resting-tremor-from-local-field-potential-recordings/. Accessed December 5, 2023.
« Back to 2018 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/an-efficient-algorithm-for-predicting-pathological-high-voltage-spindles-related-to-parkinsonian-resting-tremor-from-local-field-potential-recordings/