Embolic signal identification with novel sonographic feature and fusion algorithm for automated embolization detection/Prof. Madya Dr. Dzati Athiar Ramli
Stroke is the third leading cause of death for patients and commonly caused by embolus (plaque, blood clots, bubbles, etc) that travel to the brain anf lodge in the bloodstream. Many studies on strokes especially on automated detection of emboli signals have been reported: however, those studies only consider signal at certain topographic area of brain with certain penetration of depth during their data collection process. The inconsistency of the recorded embolic signal may happen due to it being captured from different topographic cerebral with different frequency and velocity profile. In this study , we exploit signal variability in our experiment ation by collecting emboli data from different topographic areas (MCA, PCA and ACA) at differentvelocity profiles. Two sources of inputs i.e. Doppler signal from TCD simulator and RF signal from flow phantom (in vitro) are employed. In order to resolve the variability issues, we propose a modified semi supervised technique based on expectation maximization (SEM) as it is an iterative statistical process that estimates all mixture parameters according to the highest posterior probabilities . Our methods offer several key advantages such as less processing procedures, and more robust in data variability problems compared to supervised procedure.
Embolic signal identification with novel sonographic feature and fusion algorithm for automated embolization detection/Prof. Madya Dr. Dzati Athiar Ramli
Embolic signal identification with novel sonographic feature and fusion algorithm for automated embolization detection