Showing papers on "Parametric Image published in 1998"

Parametric image reconstruction using spectral analysis of PET projection data.

Abstract: Spectral analysis is a general modelling approach that enables calculation of parametric images from reconstructed tracer kinetic data independent of an assumed compartmental structure. We investigated the validity of applying spectral analysis directly to projection data motivated by the advantages that: (i) the number of reconstructions is reduced by an order of magnitude and (ii) iterative reconstruction becomes practical which may improve signal-to-noise ratio (SNR). A dynamic software phantom with typical 2-[]thymidine kinetics was used to compare projection-based and image-based methods and to assess bias-variance trade-offs using iterative expectation maximization (EM) reconstruction. We found that the two approaches are not exactly equivalent due to properties of the non-negative least-squares algorithm. However, the differences are small and mainly affect parameters related to early and late time points on the impulse response function ( and, to a lesser extent, VD). The optimal number of EM iterations was 15-30 with up to a two-fold improvement in SNR over filtered back projection. We conclude that projection-based spectral analysis with EM reconstruction yields accurate parametric images with high SNR and has potential application to a wide range of positron emission tomography ligands.

Optimal segmentation of cell images

Abstract: An optimal segmentation algorithm for light microscopic cell images is presented. The image segmentation is performed by thresholding a parametric image approximating the original image. Using the mean squared error between the original and the constructed image as the cost function, the segmentation problem is transformed into an optimisation process where parametric parameters are determined that minimise the defined cost function. The cost function is iteratively minimised using an unsupervised learning rule to adjust the parameters, and a parametric image is constructed at each iteration, based on the obtained parameters. The cell region is extracted by thresholding the final parametric image, where the threshold is one of the image parameters. Application results to real cervical images are provided to show the performance of the proposed segmentation approach. Experimental segmentation results are presented for the proposed optimal algorithm for synthetic cell images corrupted by variant levels of noise; these results are compared with the K-means clustering method and Bayes classifier in terms of classification errors.

Method and system for correcting errors in MR images due to regions of gradient non-uniformity for parametric imaging such as quantitative flow analysis

Abstract: A system and method for correcting systematic errors that occur in MR images due to magnetic gradient non-uniformity is disclosed for use with parametric analysis. A GradWarp geometric correction operation is applied in reconstructing quantitative parametric analysis images in regions of gradient non-uniformity. The method includes generating an error map of magnetic gradient strength as a function of distance for an MR image scan and acquiring MR data that contain such systematic errors. The method next includes either calculating a measured diffusion image, a phase difference image, or similar image, based on the acquired MR data, and then calculating a corrected parametric image using the error map and the measured diffusion image, the phase difference image, or other similar parametric image. The method is incorporated into a system having a computer programmed to perform the aforementioned steps and functions.

Measurement of quantum-noise correlations in parametric image amplification

Abstract: We demonstrate quantum-noise correlations between the spatial frequencies of a parametrically amplified signal image and the generated conjugate (idler) image. Test images were amplified by an optical parametric amplifier that can be operated either as a low-pass or a band-pass amplifier for spatial frequencies. Direct difference detection of the signal and idler spatial frequencies at ±16 mm-1 resulted in noise that fell below the shot-noise level by ≃5 dB. Parametric-gain and phase-mismatch dependence of the observed quantum-noise reduction is in good agreement with the theory of a spatially-broadband optical parametric amplifier.

Fusion of cine magnetic resonance and contrast-enhanced first-pass magnetic resonance data in patients with coronary artery disease: a feasibility study.

Abstract: RATIONALE AND OBJECTIVES The left ventricle (LV) myocardial wall contractibility can be evaluated using cine magnetic resonance imaging (MRI) in a qualitative or quantitative manner. Meanwhile, myocardial perfusion can be assessed using contrast-enhanced first-pass MRI. The authors propose a method of automatically fusing the complementary information from these two cardiac MRI modalities into one single image, from which a match or mismatch between contraction and perfusion could be extracted. METHODS The authors developed a registration algorithm based on the combined use of the global affine transformation and intrinsic landmarks to match images from the same sequence or from two imaging sequences. Contraction and perfusion information was fused by combining a myocardial contour image and a parametric image of the slope of the intensity-time curve, respectively. The fusion paradigm was applied to four patients' data as a demonstration of feasibility of the proposed approach and as a preliminary evaluation. RESULTS Cine MR and contrast-enhanced MR images were well aligned. The contractibility of the LV was displayed by the myocardial contour image. The parametric slope image was consistent with the known coronary artery status of each patient. The combined contraction-perfusion representation of the LV showed the correspondence between regional LV contraction and myocardial perfusion at a one slice level. CONCLUSIONS Left ventricle contraction and myocardial perfusion can be represented conjointly in one single fused image. The fusion paradigm should be evaluated for a larger number of patients to evaluate the clinical relevance of this approach in assessing coronary artery disease.

Imagewise model fitting for generating parametric images in dynamic PET studies

Abstract: In this study, we explore the use of non-linear regression for model fitting of PET measured kinetics on a pixel-by-pixel basis for generating parametric images of micro-parameters of kinetic models. We evaluate quantitatively the noise propagation of two regression methods using computer simulated data, and examine the feasibility of generating parametric images for two different real PET studies -- a human FDG study and a monkey FDOPA study. The results demonstrated that general image-wise model fitting is practically feasible for dynamic PET studies.© (1998) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Parametric Image Reconstruction Using Spectral Analysis of (Rebinned) Three-Dimensional Projection Data

Abstract: This chapter describes methodology that enables efficient calculation of parametric images with a high signal-to-noise ratio (SNR) from three-dimensional (3D) dynamic positron emission tomography (PET) data. The 3D projections are rebinned into an equivalent 2D data set using single slice or Fourier rebinning. Spectral analysis is then applied directly to rebinned projection data, yielding projections of parameters of interest that are reconstructed using the iterative ordered subsets expectation maximization (OS-EM) algorithm. The processing time is an order of magnitude less than that for the conventional approach of fully 3D filtered backprojection (FBP) and image-based spectral analysis. The methodology was tested using the labeled anticancer drug [11C]temozolomide. Data were acquired in 3D over 90 min on an ECAT 953B tomograph (CTI/Siemens, Knoxville, TN). Parametric images proportional to flow × extraction (K1) and net rate of influx (Ki) were calculated using the new method, as well as the conventional approach. Using FBP for both image-based and projection-based methods yielded qualitatively similar results, although the new method resulted in slightly noisier images. The relative differences between the two methods, averaged over all brain voxels, were +1.6% for K1, and –2.4% for Ki. When OS-EM (1 iteration, 16 subsets) was used instead of FBP, the bias was not affected and the SNR was improved appreciably. Thus, spectral analysis of rebinned 3D projection data, combined with OS-EM reconstruction, enables efficient calculation of parametric images with high SNR.

Parametric Image Generation with Neural Networks

Abstract: An artificial neural network (ANN) is a trainable algorithm that can learn to produce an output appropriate for a given input. Such networks can be applied in a wide variety of pattern recognition tasks, including parameter estimation. The major advantages of using ANNs for parameter estimation are speed and noise tolerance. A two-layer network was used to determine the rate constants and metabolic rate (MRfdg) from dynamic PET images acquired after injection of [18F]fluorodeoxyglucose (FDG). The number of input points was 24, which represented a well-sampled tissue time-activity-curve (TAC) over 60 min after injection of FDG. One thousand noisy training data sets were generated, using the arterial plasma TAC, for each subject (computer time: ≈ 1 min). The network was trained with the backpropagation algorithm using the 1000 data sets (time: ≈ 15 min). Images were generated using the weights determined by training (time: ≈ 7 sec). Times are for a Macintosh 7100/80. The MRfdg images were of high quality (better contrast than the integrated images and less noisy than Patlak images). Parametric images (K1, k2, or k3) were quite noisy. Training the ANN with an average plasma TAC instead of the individual TACs resulted in MRfdg images of equal quality. Preliminary data from four subjects suggest it may be possible to generate quantitatively accurate images by scaling with the count from a single blood sample taken 30 min after injection.