Published online by Cambridge University Press: 14 March 2007
A procedure is presented which allows to compute in anon-invasive manner, blood viscosity through flow measurementsobtained at a fixed vessel cross-section. The data set is made ofmeasurements (artery radius and spatially discrete velocity profiles)performed at given time intervals for which the signal to noise ratiois typical of U.S. Doppler velocimetry in clinical situation. Thisidentification approach is based on the minimization, through abackpropagation algorithm, of a cost function quantifying the distancebetween numerical data obtained through Navier-Stokes simulations andexperimental measurements. Since this cost function implicitlydepends on the value of viscosity used in numerical simulations, itsminimization determines an effective viscosity which is shown to berobust to measurement errors and sampling time. Such an approach isshown to work in an in vitro experiment, and seems to besuitable for in vivo measurements of viscosity by the atraumatictechniques of Doppler echography.