Automated method for characterization of diastolic transmitral Doppler velocity contours: early rapid filling

Doppler echocardiographic studies of transmitral flow have become a routine clinical tool for the assessment and characterization of ventricular diastolic (filling) function. We have previously derived a parametrized diastolic filling (PDF) formalism for the purpose of diastolic function assessment using Doppler echocardiography. The model accommodates the mechanical "suction" feature of early diastolic filling of the heart by using a simple harmonic oscillator (SHO) as a paradigm for the kinematics of filling. PDF model predictions of transmitral flow velocity have shown excellent agreement with human echocardiographic Doppler contours (temporal profiles) when a visual, transparency overlay method of model fit to clinical Doppler contour comparison was used. The determination of PDF model parameters from the clinical Doppler contour is equivalent to the solution of the "inverse problem" of diastole. Previously, this determination consisted of a manual, iterative method of graphical overlay, in which model predicted contours were visually compared with the echocardiography machine generated Doppler contour using transparencies. To automate the process of model parameter estimation (i.e., solution of the "inverse problem") for the early or "rapid filling" phase of diastole (known in cardiology as the E-wave of the clinical Doppler velocity profile [DVP]) we recorded the acoustic pulsed Doppler signal using the forward channel of a commercial echocardiography machine. The Doppler spectrogram for a particular E-wave was recreated using short-time Fourier transform processing. The maximum velocity envelope (MVE) was extracted from the spectrogram. The PDF model was fit to the E-wave MVE using a Levenberg-Marquardt (iterative) algorithm by the requirement that the mean-square error between the clinical data (MVE) and the model be minimized. Because the model is linear, all of the PDF parameters for the Doppler E-wave can be uniquely determined. We show that: (1) solution of the "inverse problem of diastole" is possible; (2) clinical Doppler E-wave contours can be accurately reproduced and quantified using the PDF formalism and its parameters; and (3) our proposed, automated method of PDF parameter determination for the E-wave is robust.