TY - GEN
T1 - Prediction of CT Images from PET Images Using Deep Learning Approach for Small Animal Systems
AU - Nakanishi, Kouhei
AU - Yamamoto, Seiichi
AU - Watabe, Tadashi
N1 - Funding Information:
Manuscript received December 1, 2021. This work was supported in part by JSPS KAKENHI Grant Number 19H00672. Kouhei Nakanishi is with Nagoya University Graduate School of Medicine, Nagoya, Japan (e-mail: nakanishi.kouhei@c.mbox.nagoya-u.ac.jp). Seiichi Yamamoto is with Nagoya University Graduate School of Medicine, Nagoya, Japan (e-mail: s-yama@met.nagoya-u.ac.jp). Tadashi Watabe is with Osaka University Graduate School of Medicine, Osaka, Japan
Publisher Copyright:
© 2021 IEEE.
PY - 2021
Y1 - 2021
N2 - Positron emission tomography (PET) is a powerful tool because it can acquire quantitative functional images. To obtain quantitative images, attenuation correction (AC) is indispensable, but it sometimes fails in cases such as CT system problems or other causes. PET images not having CT images also appear in measurements using small animal PET systems that are not combined with a CT system or a PET/MRI system. In this case, the generation of CT images from PET images using deep learning (DL) may be a possible solution. Consequently we tried this approach using measured small animal PET/CT images. We used pix2pix generative adversarial networks (GANs) for deep learning. After training the neural network using some of the measured small animal PET/CT image pairs, we predicted synthetic CT (sCT) images from the PET images of rat heads and compared them with the measured CT images. After the training, we could generate sCT images that had similar structures to the rat's skull, although there were some differences observed in the headrest parts. We conclude that sCT image generation from PET images is possible and has the potential to be used for AC in small animal PET systems.
AB - Positron emission tomography (PET) is a powerful tool because it can acquire quantitative functional images. To obtain quantitative images, attenuation correction (AC) is indispensable, but it sometimes fails in cases such as CT system problems or other causes. PET images not having CT images also appear in measurements using small animal PET systems that are not combined with a CT system or a PET/MRI system. In this case, the generation of CT images from PET images using deep learning (DL) may be a possible solution. Consequently we tried this approach using measured small animal PET/CT images. We used pix2pix generative adversarial networks (GANs) for deep learning. After training the neural network using some of the measured small animal PET/CT image pairs, we predicted synthetic CT (sCT) images from the PET images of rat heads and compared them with the measured CT images. After the training, we could generate sCT images that had similar structures to the rat's skull, although there were some differences observed in the headrest parts. We conclude that sCT image generation from PET images is possible and has the potential to be used for AC in small animal PET systems.
UR - http://www.scopus.com/inward/record.url?scp=85139167337&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85139167337&partnerID=8YFLogxK
U2 - 10.1109/NSS/MIC44867.2021.9875591
DO - 10.1109/NSS/MIC44867.2021.9875591
M3 - Conference contribution
AN - SCOPUS:85139167337
T3 - 2021 IEEE Nuclear Science Symposium and Medical Imaging Conference Record, NSS/MIC 2021 and 28th International Symposium on Room-Temperature Semiconductor Detectors, RTSD 2022
BT - 2021 IEEE Nuclear Science Symposium and Medical Imaging Conference Record, NSS/MIC 2021 and 28th International Symposium on Room-Temperature Semiconductor Detectors, RTSD 2022
A2 - Tomita, Hideki
A2 - Nakamura, Tatsuya
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2021 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2021
Y2 - 16 October 2021 through 23 October 2021
ER -