Granted Patent B2

US 12,488,435 · App. 18/315,946 · Granted Dec 2, 2025

Inventor: Scott David Wollenweber (Waukesha, WI)

Assignee: GE PRECISION HEALTHCARE LLC

G06T5/70A61B6/032A61B6/4291G06T11/005

Abstract

Claims (60)

1 . A method for estimating noise equivalent counts, comprising:

one or more times during a scan of an object with a positron emission tomography (PET) scanner, a plurality of coincidence events are detected by a detector array of the PET scanner:

obtaining, via a processor, a total of the plurality of coincidence events detected by the detector array;

estimating, via the processor, random coincidence events detected by the detector array;

estimating, via the processor, scatter coincidence events detected by the detector array;

estimating, via the processor, true coincidence events detected by the detector array based on the total of the plurality of coincidence events, the estimated random coincidence events, and the estimated scatter coincidence events;

determining, via the processor, a scatter fraction based on the estimated scatter events and the estimated true coincidence events;

estimating, via the processor, the noise equivalent counts based at least on the scatter fraction, the total of the plurality of coincidence events, the estimated true coincidence events, the estimated scatter coincidence events, and the estimated random coincidence events; and

obtaining, via the processor, a pre-computed mask of the object imaged during the scan, wherein the pre-computed mask is in sinogram space.

2 . The method of claim 1 , wherein the pre-computed mask was obtained from a computed tomography scan of the object.

3 . The method of claim 2 , wherein determining the estimated random coincidence events that have passed through the object during the scan comprises applying the pre-computed mask to the random sinogram.

4 . The method of claim 3 , wherein determining the estimated scatter coincidence events that have passed through the object during the scan comprises applying the pre-computed mask to the scatter sinogram.

5 . The method of claim 4 , further comprising downsampling, via the processor, both the random sinogram and scatter sinogram prior to determining the estimated random coincidence events and the estimated scatter coincidence events that have passed through the object during the scan.

6 . The method of claim 1 , wherein the pre-computed mask was obtained from processed PET attenuation correction images of the object obtained by converting magnetic resonance (MR) images of the object acquired using MR sequences including zero echo time and Dixon.

7 . The method of claim 1 , further comprising, during the scan:

generating, via the processor, a random sinogram based on the estimated random coincidence events;

generating, via the processor, a scatter sinogram based on the estimated scatter coincidence events;

determining, via the processor, the estimated random coincidence events that have passed through the object during the scan based at least on the random sinogram; and

determining, via the processor, the estimated scatter coincidence events that have passed through the object during the scan based at least on the scatter sinogram.

8 . The method of claim 7 , wherein estimating the noise equivalent counts comprises estimating the noise equivalent counts based on the scatter fraction, the total of the plurality of coincidence events, the estimated random coincidence events, the estimated true coincidence events, the estimated scatter coincidence events that have passed through the object during the scan, and the estimated random coincidence events that have passed through the object during the scan.

9 . The method of claim 7 , further comprising, during the scan:

generating, via the processor, an emission sinogram based on the plurality of coincidence events.

10 . The method of claim 9 , wherein estimating scatter coincidence events comprises:

inputting, via the processor, the emission sinogram and an attenuation correction factor sinogram into a trained deep learning-based scatter estimation model; and

outputting, via the processor, from the deep learning-based scatter estimation model the scatter sinogram.

11 . The method of claim 10 , further comprising downsampling the emission sinogram and the attenuation correction factor sinogram prior to input into the deep learning-based scatter estimation model, and wherein the scatter sinogram is down-sampled.

12 . The method of claim 10 , further comprising training, via the processor, a neural network, via supervised learning, based on inputs of a target of scatter estimates and both emission sinograms and attenuation correction factor sinograms obtained from clinical examination data and/or phantom data to generate the trained deep learning-based scatter estimation model.

13 . The method of claim 1 , wherein the PET scanner comprises a plurality of detector rings disposed along a longitudinal axis of the PET scanner, and each detector ring comprises a plurality of detectors.

14 . A system for estimating noise equivalent counts, comprising:

a memory encoding processor-executable routines;

a processor configured to access the memory and to execute the processor-executable routines, wherein the processor-executable routines, when executed by the processor, cause the processor to:

one or more times during a scan of an object with a positron emission tomography (PET) scanner, a plurality of coincidence events are detected by a detector array of the PET scanner:

obtain a total of the plurality of coincidence events detected by the detector array;

estimate random coincidence events detected by the detector array;

estimate scatter coincidence events detected by the detector array;

estimate true coincidence events detected by the detector array based on the total of the plurality of coincidence events, the estimated random coincidence events, and the estimated scatter coincidence events;

determine a scatter fraction based on the estimated scatter events and the estimated true coincidence events;

estimate noise equivalent counts based at least on the scatter fraction, the total of the plurality of coincidence events, the estimated true coincidence events, the estimated scatter coincidence events, and the estimated random coincidence events; and

obtain a pre-computed mask of the object imaged during the scan, wherein the pre-computed mask is in sinogram space.

15 . The system of claim 14 , wherein the processor-executable routines, when executed by the processor, cause the processor, during the scan, to:

generate a random sinogram based on the estimated random coincidence events;

generate a scatter sinogram based on the estimated scatter coincidence events;

determine the estimated random coincidence events that have passed through the object during the scan by applying the pre-computed mask to the random sinogram;

determine the estimated scatter coincidence events that have passed through the object during the scan by applying the pre-computed mask to the scatter sinogram; and

generate an emission sinogram based on the plurality of coincidence events.

16 . The system of claim 15 , wherein estimating the noise equivalent counts comprises estimating the noise equivalent counts based on the scatter fraction, the total of the plurality of coincidence events, the estimated random coincidence events, the estimated true coincidence events, the estimated scatter coincidence events that have passed through the object during the scan, and the estimated random coincidence events that have passed through the object during the scan.

17 . The system of claim 16 , wherein estimating scatter coincidence events comprises:

inputting the emission sinogram and an attenuation correction factor sinogram into a trained deep learning-based scatter estimation model; and

outputting from the deep learning-based scatter estimation model the scatter sinogram.

18 . A non-transitory computer-readable medium, the non-transitory computer-readable medium comprising processor-executable code that when executed by a processor, causes the processor to:

one or more times during a scan of an object with a positron emission tomography (PET) scanner, a plurality of coincidence events are detected by a detector array of the PET scanner:

obtain a total of the plurality of coincidence events detected by the detector array;

generate an emission sinogram;

estimate random coincidence events detected by the detector array and generate a random sinogram based on the estimated random coincidence events;

estimate scatter coincidence events detected by the detector array by inputting the emission sinogram and an attenuation correction factor sinogram into a trained deep learning-based scatter estimation model and outputting from the deep learning-based scatter estimation model a scatter sinogram;

determine the estimated random coincidence events that have passed through the object during the scan by applying a pre-computed mask of the object to the random sinogram, wherein the pre-computed mask is in sinogram space;

determine the estimated scatter coincidence events that have passed through the object during the scan by applying the pre-computed mask to the scatter sinogram;

estimate true coincidence events detected by the detector array based on the total of the plurality of coincidence events, the estimated random coincidence events, and the estimated scatter coincidence events;

determine a scatter fraction based on the estimated scatter events and the estimated true coincidence events; and

estimate the noise equivalent counts based on the scatter fraction, the total of the plurality of coincidence events, the estimated random coincidence events, the estimated true coincidence events, the estimated scatter coincidence events that have passed through the object during the scan, and the estimated random coincidence events that have passed through the object during the scan.

Assignments (2)

Continuity (1)

References Cited (16)