Abstract

Moustafa Hussein Aly
Robustness to noise test for the machine learning model of neurology problems
Cerebral Vasospasm (CV) is a narrowing of the blood vessels in thehuman brain. Transcranial Doppler (TCD) is anoninvasive device and can be usedas for diagnosis various brain diseases and CV detection. TCD signals can be contaminated with noise from sources power line and electrodes before using these signals in signal processing steps. The goal of this study is to evaluate the CV detection model accuracy against the noise. Time-frequency feature extraction wasisused as a technique to enhance the detection accuracy and efficiency. In theprevious studies, we extracted CV and normal classifier model by using the combination of 12-time frequency features, but the results generateda moderate accuracy when examined in real-time[put a reference]. In this study, we test the robustness to noise of proposed model experimentswereareapplied in real-time on the recorded TCD signal from theright and left middle cerebral artery (MCA) region of thebrain of 160 subjects. The experimental results giveus87.5% sensitivity for CV. Thispercentage starts to decrease at 30%???of signal to noise ratio (SNR), and 89% specificity for normaland this percentage starts to decrease at 60%??? of SNR.