Design And Application Of A Stroke Classification And Prediction System Based On Peripheral LFO

In medical imaging,researchers have used various medical imaging techniques to analyze the characteristics of ischemic stroke,determine the occurrence of stroke and make a prognosis.Although magnetic resonance imaging and computed tomography can provide accurate results for ischemic stroke diagnosis,they are expensive and their use is limited.There is a lack of a method and device that can continuously detect blood flow changes in blood vessels that cause ischemic stroke over a long period of time.To solve this challenge,the low-frequency oscillations(LFO)signal was applied to ischemic stroke for system development,feature extraction and classification prediction.First,a stroke classification and prediction system based on peripheral LFO was developed to address the current problem of lacking a portable and rapid assessment system for stroke diagnosis.The limitations of the traditional NIR system working mode were improved,and the working mode of this system was designed.Based on this,the performance characteristics of the light source were analyzed,and the light source model and the optimal wavelength were determined,and then the emitting diode driving circuit was designed.The performance of the detector is analyzed,and the signal detection module circuit is designed to complete the signal amplification and filtering processing.To avoid crosstalk between multiple channels,a multi-channel signal synchronization circuit is designed to realize real-time synchronized acquisition of human peripheral LFO signals.Second,the stroke classification prediction system was applied to clinical signal acquisition experiments and feature extraction to address the problem of the inability to continuously and dynamically detect ischemic stroke peripheral blood flow propagation in medical research.After the system was developed and tested for performance,it was applied to healthy subjects and ischemic stroke patients for resting-state peripheral LFO signal acquisition,and an ensemble empirical modal decomposition based on correlation coefficients(referred to as x Corr-EEMD)algorithm is used to remove the signal motion noise..Then,time-frequency domain features were extracted from the peripheral LFO signals of the subjects to construct baseline criteria for peripheral LFO assessment of ischemic stroke,and the distribution of baseline deviation characteristics of ischemic stroke patients was further analyzed.Finally,a stroke classification prediction model was constructed and validated.The extracted significant difference features between ischemic stroke patients and healthy subjects were trained by applying a random forest algorithm for classification to find out the best predictive features,and ROC curves were plotted to verify the accuracy of classification prediction;further classification of ischemic stroke according to the degree of embolism was performed to verify the applicability of stroke prediction based on peripheral LFO;finally,clinical trials were conducted on four patients with atherosclerosis to verify the reliability of this classification prediction model.