Direct reconstruction for simultaneous dual-tracer PET imaging based on multi-task learning

Abstract

Background: Simultaneous dual-tracer positron emission tomography (PET) imaging can observe two molecular targets in a single scan, which is conducive to disease diagnosis and tracking. Since the signals emitted by different tracers are the same, it is crucial to separate each single tracer from the mixed signals. The current study proposed a novel deep learning-based method to reconstruct single-tracer activity distributions from the dual-tracer sinogram.

Methods: We proposed the Multi-task CNN, a three-dimensional convolutional neural network (CNN) based on a framework of multi-task learning. One common encoder extracted features from the dual-tracer dynamic sinogram, followed by two distinct and parallel decoders which reconstructed the single-tracer dynamic images of two tracers separately. The model was evaluated by mean squared error (MSE), multiscale structural similarity (MS-SSIM) index and peak signal-to-noise ratio (PSNR) on simulated data and real animal data, and compared to the filtered back-projection method based on deep learning (FBP-CNN).

Results: In the simulation experiments, the Multi-task CNN reconstructed single-tracer images with lower MSE, higher MS-SSIM and PSNR than FBP-CNN, and was more robust to the changes in individual difference, tracer combination and scanning protocol. In the experiment of rats with an orthotopic xenograft glioma model, the Multi-task CNN reconstructions also showed higher qualities than FBP-CNN reconstructions.

Conclusions: The proposed Multi-task CNN could effectively reconstruct the dynamic activity images of two single tracers from the dual-tracer dynamic sinogram, which was potential in the direct reconstruction for real simultaneous dual-tracer PET imaging data in future.

Keywords: Dual-tracer PET; Multi-task learning; Reconstruction; Signal separation.

Fig. 1

The architecture of Multi-task CNN. The dual-tracer dynamic sinogram is inputted to the encoder, and two decoders output the dynamic activity images of two tracers. Conv = convolution layer, BN = batch normalization layer, ReLU = rectified linear unit layer, Deconv=deconvolution layer. The kernel size and number of channels are noted in the figure

Fig. 2

The Zubal phantoms used in simulation experiments. a b, c, and d are representative phantoms corresponding to differenct slices of the original 3D phantom

Fig. 3

The representative reconstructed images from the individual difference experiment. a 10% kinetic variation, b 20% kinetic variation. These images show the 10th frame of the dynamic images

Fig. 4

The representative reconstructed images from the tracer combination experiment. a Tracer combination I (${}^{11}$C-FMZ/${}^{11}$C-acetate), b Tracer combination II (${}^{18}$F-FDG/${}^{11}$C-FMZ). These images show the 10th frame of the dynamic images

Fig. 5

The representative reconstructed images from Protocol III (60-min scan) of the scanning protocol experiment

Fig. 6

The ROI-TACs from the scanning protocol experiment

Fig. 7

The representative reconstructed images from the animal experiment. These images show the same slice of one rat