Motion correction of respiratory-gated PET images using deep learning based image registration framework

Abstract

Artifacts caused by patient breathing and movement during PET data acquisition affect image quality. Respiratory gating is commonly used to gate the list-mode PET data into multiple bins over a respiratory cycle. Non-rigid registration of respiratory-gated PET images can reduce motion artifacts and preserve count statistics, but it is time consuming. In this work, we propose an unsupervised non-rigid image registration framework using deep learning for motion correction. Our network uses a differentiable spatial transformer layer to warp the moving image to the fixed image and uses a stacked structure for deformation field refinement. Estimated deformation fields were incorporated into an iterative image reconstruction algorithm to perform motion compensated PET image reconstruction. We validated the proposed method using simulation and clinical data and implemented an iterative image registration approach for comparison. Motion compensated reconstructions were compared with ungated images. Our simulation study showed that the motion compensated methods can generate images with sharp boundaries and reveal more details in the heart region compared with the ungated image. The resulting normalized root mean square error (NRMS) was 24.3 ± 1.7% for the deep learning based motion correction, 31.1 ± 1.4% for the iterative registration based motion correction, and 41.9 ± 2.0% for ungated reconstruction. The proposed deep learning based motion correction reduced the bias compared with the ungated image without increasing the noise level and outperformed the iterative registration based method. In the real data study, both motion compensated images provided higher lesion contrast and sharper liver boundaries than the ungated image and had lower noise than the reference gate image. The contrast of the proposed method based on the deep neural network was higher than the ungated image and iterative registration method at any matched noise level.

Deep Learning Based Joint PET Image Reconstruction and Motion Estimation

“…In this study, we apply our previously developed unsupervised deep learning model for deformable motion estimation [21]. The overall network architecture is shown in Fig.…”

“…In order to assign a good initialization for 𝜽 𝑚 , we use the gated Maximum Likelihood Expectation Maximization (ML-EM) reconstructed images to estimate the initial values of 𝜽 𝑚 . For 𝜶 𝑚 and 𝜷 initialization, we first obtained a motion compensated reconstruction 𝒙 𝑖𝑛𝑖 [21], and used 𝒙 𝑖𝑛𝑖 to initialize 𝜷 and warped 𝒙 𝑖𝑛𝑖 based on 𝜽 𝑚 to initialize 𝜶 𝑚 .…”

“…Recently, deep learning (DL) techniques have provided new approaches to image registration [14]- [19]. There are three widely used components of DL-based image registration [20], an encoder-decoder based architecture incorporating several hierarchical convolution layers for multi-scale feature extraction [16], [19], [21], a spatial transformer network (STN) [22] for spatial transformation, and a generative adversarial network (GAN) [23]- [25] where a generator predicts the deformation field and warps the moving image and the warped moving image is evaluated by a discriminator. One advantage of DL methods is that a pretrained network can incorporate prior knowledge of motion fields from training data into the motion estimation process.…”

“…One advantage of DL methods is that a pretrained network can incorporate prior knowledge of motion fields from training data into the motion estimation process. Previously, we proposed an unsupervised non-rigid image registration framework using DL to estimate deformation fields for respiratory gated images [21] and demonstrated that it could outperform an iterative image registration method.…”

MDPET: A Unified Motion Correction and Denoising Adversarial Network for Low-Dose Gated PET

“…The motion vectors derived from the image registration were then utilized to average transformed images or incorporated into a final reconstruction to generate a motion compensated image with all events. In addition to using the conventional non-rigid image registration algorithms [4]- [7], deep learning based methods were explored recently as well [8], [9]. However, the noisy gated images could lead to inaccurate motion estimation and alignment errors.…”