US 11,756,241 B2
Positron emission tomography system and image reconstruction method using the same
Jae Sung Lee, Seoul (KR); Donghwi Hwang, Seoul (KR); and Kyeong Yun Kim, Seoul (KR)
Assigned to SEOUL NATIONAL UNIVERSITY R&DB FOUNDATION, Seoul (KR)
Filed by Seoul National University R&DB Foundation, Seoul (KR)
Filed on Feb. 8, 2021, as Appl. No. 17/169,687.
Application 17/169,687 is a continuation in part of application No. 16/282,513, filed on Feb. 22, 2019, granted, now 10,943,349.
Claims priority of application No. 10-2018-0021902 (KR), filed on Feb. 23, 2018.
Prior Publication US 2021/0166444 A1, Jun. 3, 2021
This patent is subject to a terminal disclaimer.
Int. Cl. G06T 11/00 (2006.01); G06T 3/40 (2006.01); A61B 6/03 (2006.01); A61B 6/00 (2006.01); G01T 1/29 (2006.01)
CPC G06T 11/005 (2013.01) [A61B 6/037 (2013.01); A61B 6/5258 (2013.01); G01T 1/2985 (2013.01); G06T 3/40 (2013.01); G06T 2210/41 (2013.01)] 13 Claims
OG exemplary drawing
 
1. A positron emission tomography system comprising:
a collection unit collecting a positron emission tomography sinogram (PET sinogram);
a generation unit applying a maximum likelihood reconstruction of attenuation and activity (MLAA) with time-of-flight (TOF) to the positron emission tomography sinogram and generating a first emission image and a first attenuation image, and a nonattenuation-corrected (NAC) image reconstructed without attenuation correction;
a control unit selecting at least one of the first emission image and the first attenuation image and the NAC image generated by the generation unit as an input image and generating and providing a final attenuation image by applying the learned deep learning algorithm to the input image; and
a learning unit collecting an attenuation image obtained through additional scanning based on the positron emission tomography sinogram and making a deep learning algorithm be learned by using at least one of the first emission image, the first attenuation image and the NAC image generated from the positron emission tomography sinogram and the obtained attenuation image, wherein:
the learning unit includes,
an image generation unit generating a second attenuation image from the first attenuation image through the deep learning algorithm,
an error calculation unit calculating an error between the second attenuation image and the obtained attenuation image, and
a weight adjustment unit performing repeated learning by readjusting weights of a plurality of filters included in the deep learning algorithm so as to generate a final attenuation image in which the error value becomes a value equal to or less than a threshold value.