Features Extraction For Parkinson's Disease In Local Field Potentials

Deep brain stimulation(DBS)is an effective treatment for advanced Parkinson's disease.However,the current DBS system is difficult to adapt to the changes of patients' physiological status,and it also lacks of self-adaptability and has side effects.Currently,researchers pay more attention on the adaptive DBS system.The ideal adaptive DBS system can adaptively control the stimulation parameters according to the physiological status of patients.Obtaining biomarkers that can reflect the physiological status of patients is the core of adaptive DBS system research,and it is also an urgent problem to be solved at present.The local field potential(LFP)of Parkinson's disease patients show abnormal oscillation.Extracting abnormal oscillation signals is useful for the pathological mechanism of Parkinson's disease patients.Designing the stimulation control rules of adaptive DBS has great significance for the treatment of late Parkinson's disease patients.In this paper,we studied the LFP signals of 8 patients with Parkinson's disease.In order to obtain the biomarkers of Parkinson's disease,we have done the following work:Firstly,this paper has summarized the current feature extraction methods of Parkinson's disease,analyzed the principle of the algorithm and the performance of biomarkers obtained,and proposed the design criteria of the feature extraction algorithm for Parkinson's disease;Secondly,the LFP signals of 8 patients with Parkinson's disease were preprocessed and analyzed in spectral to obtain the frequency domain characteristics of abnormal oscillation.Among them,recorded(stimulation group)LFP signals of two patients while stimulating with DBS,and cut off six patients'(unstimulated group)medications to record the signals of local field potentials in resting state.The spectrum analysis shows that the frequency band of the abnormal oscillation signal is 8-35 Hz.Thirdly,an improved empirical mode algorithm is designed to extract the abnormal oscillation signal,and the abnormal oscillation is decomposed from the time domain.A feature extraction algorithm based on signal shaping,waveform counting and threshold envelope was designed,which from the angle of accuracy and computational efficiency.Fourthly,the obtained abnormal oscillation were analyzed statistically and compared with the existing fixed threshold method,the decomposed abnormal oscillation signal contains more effiective information,the accuracy of the algorithm is verified by this result;and compared with the non-stimulated group,the stimulated group has less long oscillation and more short oscillation,the comparison result also can verify the effectiveness of the algorithm.This paper aim to obtain the biological markers of Parkinson's disease.The existing algorithms are analyzed and the design requirements of Parkinson's disease feature extraction algorithm has been made in this paper,at the same time,an algorithm from the time domain decomposition of Parkinson's disease patients with local field potential abnormal oscillation signal has been designed,and,through the validation analysis,the accuracy and effectiveness of the algorithm were verified in this paper,finally,this paper had prepared for the next step of adaptive DBS system design work.