Respiratory Motion Estimation With Facilitation of a Physiology-Informed Deep Convolutional Denoising Network for 18F-FDG PET

Résumé fourni par la source

The presence of respiratory motion during PET acquisition degrades the reconstructed image. Although gating methods that sort the data into several gates according to the breathing phase or displacement help reduce the effect, the higher noise level due to less counts in the gated images can hamper the diagnostic accuracy and/or subsequent motion estimation among the gates. Recently, a physiology-informed deep convolutional neural network (PI-DCNN) specifically designed for 18F-FDG PET image denoising is proposed. In this study, we investigate the feasibility of using the network to facilitate the motion estimation hence compensation among the gates. Three cancer patient datasets acquired on a Canon Cartesion PET/CT scanner with Anzai signals recorded during the scan for gating are included. Preliminary results demonstrate that smoother and more consistent motion fields among the gates can be obtained while the PI-DCNN is utilized for gated image denoising. This in turn improves the image sharpness and structure delineation in the final motion compensated image.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.