STATISTICAL AND NEURO-FUZZY APPROACHES FOR EMBOLI DETECTION (FriAmPO4)

Author(s) : Denis Kouame (LUSSI CNRS FRE 2448, FRANCE) Mathieu Biard (LUSSI CNRS FRE 2448, FRANCE) Jean-Marc Girault (LUSSI CNRS FRE 2448, FRANCE) Aurore Bleuzen (LUSSI CNRS FRE 2448, FRANCE) François Tranquart (LUSSI CNRS FRE 2448, FRANCE) Frédéric Patat (LUSSI CNRS FRE 2448, FRANCE)

Abstract : Relation between cerebral emboli occurrence and stroke has been suggested these last years. Emboli detection has then become a constant concern while monitoring cerebral vascular pathologies. This detection is based on analysis of embolic TransCranial Doppler (TCD) signal. In practical experiments, most of detected emboli are big-size emboli ones, because of their easy-to-recognize signature in the TCD signal. The problem of small size emboli detection is an opened one and remains a challenge. Different approaches have been proposed to solve this problem. They use exclusively human expert knowledge or automatic collection of signal parameters. In this paper we propose to used both expert knowledge and automatic processing through neuro-fuzzy approach. Performances evaluation and comparison with high performance micro-emboli detection technique, namely Autoregressive (AR) modelling are provided, using \textit{in vitro} in this work.

Menu