Title:
MEDICAL IMAGE PROCESSOR AND PROGRAM
Document Type and Number:
Japanese Patent JP2019211475
Kind Code:
A
Abstract:
To improve image quality of a medical image processor.SOLUTION: A medical image processor includes an acquisition section and a processing section. The acquisition section acquires a PET (Positron Emission Tomography) image acquired on the basis of detecting a plurality of detection elements and reconstituted from emission data indicating simultaneous measurement of each pair of gamma rays generated from a pair annihilation event. The processing section outputs the PET image denoised by inputting the PET image reconstituted by a reconstitution section for a learned model outputting the denoised PET image by inputting the PET image. The learned model is learned by optimizing a loss function so that difference between a reconstitution image of a first noise level and reconstitution images of a plurality of noise levels higher than the first noise level is minimized.SELECTED DRAWING: Figure 1
Inventors:
CHANG CHUN
ZHOU JIAN
EVREN ASMA
ZHOU JIAN
EVREN ASMA
Application Number:
JP2019100684A
Publication Date:
December 12, 2019
Filing Date:
May 29, 2019
Export Citation:
Assignee:
CANON MEDICAL SYSTEMS CORP
International Classes:
G01T1/161; G06T5/00; G06T7/00
Foreign References:
| WO2018039380A1 | 2018-03-01 |
Other References:
HONGMING SHAN ET AL.: "3-D Convolutional Encoder-Decoder Network for Low-Dose CT via Transfer Learning From a 2-D Trained N", IEEE TRANSACTIONS ON MEDICAL IMAGING, vol. Volume: 37, Issue: 6, JPN6023004626, 2 May 2018 (2018-05-02), pages 1522 - 1534, ISSN: 0004982357
Attorney, Agent or Firm:
Toranomon Intellectual Property Office