Identification of effective features of LFP signal for making closed-loop deep brain stimulation in parkinsonian rats

Mohammad Pooyan | Mehrdad Roghani | sana amoozegar

---

URL :   http://research.shahed.ac.ir/WSR/WebPages/Report/PaperView.aspx?PaperID=159164
Date :  2022/01/10
Publish in :    Medical and Biological Engineering and Computing
DOI :  https://doi.org/https://doi.org/10.1007/s11517-021-02470-3

Keywords :Chaotic biomarker · Closed-loop deep brain stimulation · Local feld potential · Parkinson’s disease · Support vector machine

Abstract :

---

Traditional deep brain stimulation (DBS) is one of the acceptable methods to relieve the clinical symptoms of Parkinson’s disease in its advanced stages. Today, the use of closed-loop DBS to increase stimulation efciency and patient satisfaction is one of the most important issues under investigation. The present study was aimed to fnd local feld potential (LFP) features of parkinsonian rats, which can determine the timing of stimulation with high accuracy. The LFP signals from rats were recorded in three groups of parkinsonian rat models receiving stimulation (stimulation), without getting stimulation (of-stimulation), and sham-controlled group. The frequency domain and chaotic features of signals were extracted for classifying three classes by support vector machine (SVM) and neural networks. The best combination of features was selected using the genetic algorithm (GA). Finally, the efective features were introduced to determine the on/of stimulation time, and the optimal stimulation parameters were identifed. It was found that a combination of frequency domain and chaotic features with an accuracy of about 99 was able to determine the time the DBS must switch on. In about 80.67 of the 1861 diferent stimulation parameters, the brain was able to maintain its state for about 3 min after stimulation discontinuation.