Showing papers on "Parametric Image published in 1996"

CHAPTER 57 – Generation of Parametric Images Using Factor Analysis of Dynamic PET

Abstract: The estimation of physiologic parameters from dynamic positron emission tomography is limited by the validity of the assumed kinetic model. Furthermore, the presence of noise introduces variability that can degrade the precision and accuracy of the parameter estimates. In the case of parametric imaging, the estimation can be especially inadequate because it is performed at the pixel level with much lower counts than averaged regions of interest. These limitations can be reduced or avoided by applying data-driven statistical techniques such as factor analysis to identify the relevant temporal signatures while removing the random variations due to noise. In the first case, principal component analysis (PCA) is used to condition the data prior to applying the conventional parameter estimation, thereby improving the noise limitations. In the second case, the PCA provides the initial solution for the factor analysis, which incorporates prior knowledge to generate factor images, and time factors, which themselves can be considered a form of parametric image, and their associated physiologic time functions. In this regard, data-driven parameter estimation can be performed that avoids the inconsistencies between the actual data and the assumed kinetic model. Factor analysis has already been applied to dynamic neuroreceptor and cerebral glucose metabolism positron emission tomography (PET) studies to generate meaningful factors that differentiate specific from nonspecific binding and normal from diseased tissue function, respectively. In the latter case, a simulated image sequence having a pathophysiologic disturbance is generated and analyzed by factor analysis to produce factors consistent with the clinical findings. Significant improvements can also be achieved in the estimation of the compartmental model rate constants using the PCA processed data as compared to the raw simulated sequence.

Imaging Through Scattering Media by Parametric Image Amplification

Abstract: Parametric image amplification is proposed as a new solution for detecting objects hidden by scattering media. An image of a resolution chart has been obtained through a milk cell 18 scattering mean free paths thick. The ballistic light is amplified with a mean gain of 40 dB and the resolution on the object is about 20 μm.