http://rdf.ncbi.nlm.nih.gov/pubchem/patent/KR-102296881-B1
Outgoing Links
| Predicate | Object |
|---|---|
| classificationCPCAdditional | http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06T2207-10104 |
| classificationCPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06T11-005 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/A61B6-54 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/A61B6-037 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/A61B6-5235 |
| classificationIPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/A61B6-03 http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/A61B6-00 http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06T11-00 |
| filingDate | 2019-08-28-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
| grantDate | 2021-09-02-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
| publicationDate | 2021-09-02-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
| publicationNumber | KR-102296881-B1 |
| titleOfInvention | System for reconstructing quantitative PET dynamic image using neural network and Complementary Frame Reconstruction |
| abstract | The present invention relates to a quantitative PET dynamic image reconstruction system using an artificial neural network and CFR. An artificial neural network that calculates target frame data of high SNR and temporal resolution based on CFR for dynamic PET image reconstruction, and Dynamic PET considering CFR Includes dataset generator. This Dynamic PET dataset generator selects the images corresponding to the target frame and the complementary frame from the image dataset, and matches the weight of the target frame from the total recording time of Dynamic PET to the target frame and complementary frame for each of them. a first pre-processing unit that adjusts pixel values of an image corresponding to a frame; a second preprocessor that converts the target frame and the complementary frame, each whose pixel values are adjusted in size to match the shooting time, into a sinogram; a third preprocessor that generates sinograms corresponding to the target frame (noisy) and complementary frame (noisy) with Poisson noise added to the sinograms corresponding to the target frame and the complementary frame; By adding the target frame (noisy) and complementary frame (noisy) containing Poisson noise, the full frame (noisy) corresponding to the entire shooting time of the target frame (noisy) and complementary frame (noisy) is calculated, and the full frame (noisy) is calculated. ) and a first output unit for setting the training data of the artificial neural network by generating a sinogram of 2 channels so that pixel values of complementary frames (noisy) have respective channel values; And the target frame (noisy) and complementary frame (noisy) containing Poisson noise obtained through the third preprocessor are reconstructed into an image using Filtered Back Projection and added to capture the target frame (noisy) and complementary frame (noisy). It calculates the full frame (noisy) corresponding to the entire time, and sets the training data of the artificial neural network by creating two-channel images so that the pixel values of the full frame (noisy) and the complementary frame (noisy) have each channel value. and a second output. Then, the artificial neural network receives the full frame (noisy) and complementary frame (noisy), which are two-channel input data in sinogram form, from the first output unit, extracts the main representation of the target frame, and solves the extracted target frame representation. The sinogram of the target frame without noise is calculated, and the main representation of the target frame is extracted by receiving the full frame (noisy) and complementary frame (noisy), which are two-channel input data in the form of images, from the second output unit, An image of the target frame without Poisson noise is calculated from the extracted representation of the target frame. |
| priorityDate | 2019-08-28-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
| type | http://data.epo.org/linked-data/def/patent/Publication |
Incoming Links
Total number of triples: 19.