| Summary: | 碩士 === 國立臺灣大學 === 電機工程學研究所 === 91 === Ultrasonic contrast agents have been used to enhance the acoustic backscattered intensity of blood and to assist the assessment of blood flow parameters. One example is the time-intensity method based on the indicator-dilution theory. The method has been applied to study cancer and myocardial blood flow. However, the method does not work properly for brain perfusion measurements due to the high attenuation of the transtemporal bone window. In addition, when the injection site is far away from the measurement site, the input time-intensity curve (TIC) is no longer an impulse function so that blood flow parameters cannot be obtained directly from measured TIC. In this case, the effects resulted from non-instantaneous injection (i.e., the input TIC is not an impulse function) needs to be removed before assessing blood flow parameters.
The TIC method estimates flow parameters by observing the dilution process of contrast agents in the region of interest (ROI). The compartment model has been employed to describe such a dilution process. Under the linear and time-invariant assumption, the TIC measured at the output of a compartment is the convolution of the input TIC with the compartment's transfer function. According to this property, effects of non-instantaneous injection previously mentioned can be eliminated through transfer function analysis. The transfer function analysis is possible using deconvolution when both the input and the output time intensities are available. Therefore, an input/output TIC (IOTIC) method is proposed in this study, and a deconvolution technique is employed for transfer function analysis. In addition, blood flow estimation over the perfusion area which cannot be effectively imaged by ultrasound can also benefit from the proposed deconvolution technique. Note that the linear and time-invariant assumption requires a constant flow rate. With flow pulsation, however, the flow rate changes with time and the mixing process becomes time varying. Thus, the effects of flow pulsation on the time-intensity measurements are investigated first in this study. This study also proposes deconvolution techniques based on the ensemble method, the cross-correlation method and the recursive least-squares method for time-varying transfer function analysis. Limitations and efficacy of the three deconvolution techniques are also discussed.
Both simulations and experiments are performed in this study. Results indicate that the pulsation generally does not affect the validity of time-intensity based flow estimation. The proposed RLS deconvolution technique is effective for both constant and pulsatile flows, thus permitting transfer function analysis in various flow conditions. Also, the physical meanings of the deconvolution results under flow pulsation are addressed. The efficacy of the deconvolution technique for time-varying transfer function analysis has been demonstrated in this study. Using the proposed technique, we hope to further reconstruct 2-D time-varying transfer function of ROI in the perfused area, and to further provide more useful blood flow parameters in tumor and heart, and estimate brain perfusion based on extracranial time-intensity measurements.
|
|---|
