Integrating ultrasound images using modular artificial neural networks

Abstract

The aim of the work, described in this paper, was to develop a new set-up to infer on significant disturbances of carotid artery hemodynamics based on the analysis of clinical Doppler acquisitions. A patient-specific numerical simulation system for the analysis of arterial blood flow under pulsatile conditions using a modular artificial neural network with optimal configuration was implemented based on carotid geometry and Doppler hemodynamic features. With arterial blood being addressed as a time series dependent on heart pulse periodicity, splitting parameters given by carotid bifurcation branches were able to allocate the input-output vectors for different flow regimes. Predictions given by the optimal modular artificial neural network model were able to reproduce patient-specific velocity patterns. The hemodynamic behaviour of healthy, low grade and high grade carotid artery stenosis was addressed suggesting further developments for hemodynamic classification. (c) Civil-Comp Press, 2015.