Browse Prior Art Database

ADAPTIVE MULTIPLE-SCATTER ESTIMATION METHOD FOR MODEL-BASED SCATTER CORRECTION IN POSITRON EMISSION TOMOGRAPHY

IP.com Disclosure Number: IPCOM000117352D
Publication Date: 2005-Mar-31
Document File: 6 page(s) / 40K

Publishing Venue

The IP.com Prior Art Database

Abstract

In one embodiment, using bounded values of kernel width (30 < s < 100, mm) and normalization (N, 0.1 < N < 0.3), the multiple scatter are initially calculated using a "best guess" mean value. Following generation of the multiple-scatter per slice (sinogram), each sinogram is collapsed in angle to generate a profile in r. Likewise, the input T+S emission sinogram is collapsed. The support (from the attenuation sinogram) is also collapsed to allow computation of the object boundary, hence determining the location of the tails (L/R). In the tail regions, a total-squared-error (TSE) calculation is performed by calculating the square of the difference on a point-by-point basis between the two profiles. This is to be summed over all sinograms in the dataset.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 33% of the total text.

ADAPTIVE MULTIPLE-SCATTER ESTIMATION METHOD FOR MODEL-BASED SCATTER CORRECTION IN POSITRON EMISSION TOMOGRAPHY

FIELD OF THE INVENTION

[0001]   This invention relates generally, to scatter estimation methods and more particularly, to a multiple-scatter estimation method for a model based scatter correction in positron emission tomography (PET).

BACKGROUND OF THE INVENTION

[0002]    A common method of scatter correction for fully-3D positron emission tomography (PET) is to use model-based scatter correction (MBSC).  This method uses the raw measured emission counts to iteratively calculate the theoretical scatter distribution from the imaged object.  In most whole-body imaging situations in a clinical-sized PET scanner, the contribution of scatter (S) is approximately equal to the total number of true (T) coincidence events measured (i.e., a scatter fraction of 50%).  Of these scattered events, the model only theoretically calculates the single-scatter (SS) events, defined as a coincidence event where one of the photons does not scatter but is detected and the other scatters once prior to being detected.  From simulation on body-sized objects, this type of event accounts for approximately 75-90% of all detected scatter events.  The actual contribution of multiply-scattered (MS) events depends on the object source and attenuation distribution and the scanner geometry. 

[0003]   Another known method of estimating the multiple-scatter distribution from the single-scatter distribution is to use a simple convolution of the SS with a Gaussian of defined width and normalization, then use the information outside the object to scale the SS+MS result to obtain the total scatter (TS) estimate.  The object boundary is reliably found from a measurement of object attenuation, and the assumption is that all remaining events outside the object must be due to scatter.  Prior corrections for random (R) coincidence events are assumed to be bias-free.  In this manner, the total scatter is “tail-scaled” to match the data to be corrected.  

[0004]    This method, however, has a few shortcomings.  First, any residual events in the measured emission data tails which are not scatter events will be treated as scatter, which can cause scatter scaling problems.  Further, the current implementation includes a set multiple-scatter convolution kernel width and normalization, which is not rigorously correct for any object but represents the average minimum-squared-error found from simulation over a variety of objects.  Given that the multiple-scatter varies from object to object but is not treated as such, the algorithm will be sensitive to exact data inclusion/exclusion in the tails used for scaling.  To this end, one would prefer to minimize the dependence on the tail-scaling. 

[0005]    Thus, there exists a need for a multiple-scatter estimation method, for a model based scatter correction PET, wherein (i) the dependence of the final scatter estimate on the fu...