Low Dose PET Image Reconstruction with Total Variation Using Alternating Direction Method

Employing Multiple Low-Dose PET Images (at Different Dose Levels) as Prior Knowledge to Predict Standard-Dose PET Images

The existing deep learning-based denoising methods predicting standard-dose PET images (S-PET) from the low-dose versions (L-PET) solely rely on a single-dose level of PET images as the input of deep learning network. In this work, we exploited the prior knowledge in the form of multiple low-dose levels of PET images to estimate the S-PET images. To this end, a high-resolution ResNet architecture was utilized to predict S-PET images from 6 to 4% L-PET images. For the 6% L-PET imaging, two models were developed; the first and second models were trained using a single input of 6% L-PET and three inputs of 6%, 4%, and 2% L-PET as input to predict S-PET images, respectively. Similarly, for 4% L-PET imaging, a model was trained using a single input of 4% low-dose data, and a three-channel model was developed getting 4%, 3%, and 2% L-PET images. The performance of the four models was evaluated using structural similarity index (SSI), peak signal-to-noise ratio (PSNR), and root mean square error (RMSE) within the entire head regions and malignant lesions. The 4% multi-input model led to improved SSI and PSNR and a significant decrease in RMSE by 22.22% and 25.42% within the entire head region and malignant lesions, respectively. Furthermore, the 4% multi-input network remarkably decreased the lesions' SUVmean bias and SUVmax bias by 64.58% and 37.12% comparing to single-input network. In addition, the 6% multi-input network decreased the RMSE within the entire head region, within the lesions, lesions' SUVmean bias, and SUVmax bias by 37.5%, 39.58%, 86.99%, and 45.60%, respectively. This study demonstrated the significant benefits of using prior knowledge in the form of multiple L-PET images to predict S-PET images.

Delayed FDG PET Provides Superior Glioblastoma Conspicuity Compared to Conventional Image Timing

Background: Glioblastomas are malignant, often incurable brain tumors. Reliable discrimination between recurrent disease and treatment changes is a significant challenge. Prior work has suggested glioblastoma FDG PET conspicuity is improved at delayed time points vs. conventional imaging times. This study aimed to determine the ideal FDG imaging time point in a population of untreated glioblastomas in preparation for future trials involving the non-invasive assessment of true progression vs. pseudoprogression in glioblastoma.

Methods: Sixteen pre-treatment adults with suspected glioblastoma received FDG PET at 1, 5, and 8 h post-FDG injection within the 3 days prior to surgery. Maximum standard uptake values were measured at each timepoint for the central enhancing component of the lesion and the contralateral normal-appearing brain.

Results: Sixteen patients (nine male) had pathology confirmed IDH-wildtype, glioblastoma. Our results revealed statistically significant improvements in the maximum standardized uptake values and subjective conspicuity of glioblastomas at later time points compared to the conventional (1 h time point). The tumor to background ratio at 1, 5, and 8 h was 1.4 ± 0.4, 1.8 ± 0.5, and 2.1 ± 0.6, respectively. This was statistically significant for the 5 h time point over the 1 h time point (p > 0.001), the 8 h time point over the 1 h time point (p = 0.026), and the 8 h time point over the 5 h time point (p = 0.036).

Conclusions: Our findings demonstrate that delayed imaging time point provides superior conspicuity of glioblastoma compared to conventional imaging. Further research based on these results may translate into improvements in the determination of true progression from pseudoprogression.

LCPR-Net: low-count PET image reconstruction using the domain transform and cycle-consistent generative adversarial networks

BACKGROUND

Reducing the radiation tracer dose and scanning time during positron emission tomography (PET) imaging can reduce the cost of the tracer, reduce motion artifacts, and increase the efficiency of the scanner. However, the reconstructed images to be noisy. It is very important to reconstruct high-quality images with low-count (LC) data. Therefore, we propose a deep learning method called LCPR-Net, which is used for directly reconstructing full-count (FC) PET images from corresponding LC sinogram data.

METHODS

Based on the framework of a generative adversarial network (GAN), we enforce a cyclic consistency constraint on the least-squares loss to establish a nonlinear end-to-end mapping process from LC sinograms to FC images. In this process, we merge a convolutional neural network (CNN) and a residual network for feature extraction and image reconstruction. In addition, the domain transform (DT) operation sends a priori information to the cycle-consistent GAN (CycleGAN) network, avoiding the need for a large amount of computational resources to learn this transformation.

RESULTS

The main advantages of this method are as follows. First, the network can use LC sinogram data as input to directly reconstruct an FC PET image. The reconstruction speed is faster than that provided by model-based iterative reconstruction. Second, reconstruction based on the CycleGAN framework improves the quality of the reconstructed image.

CONCLUSIONS

Compared with other state-of-the-art methods, the quantitative and qualitative evaluation results show that the proposed method is accurate and effective for FC PET image reconstruction.

Improved discrimination between benign and malignant LDCT screening-detected lung nodules with dynamic over static 18F-FDG PET as a function of injected dose

Lung cancer mortality rate can be significantly reduced by up to 20% through routine low-dose computed tomography (LDCT) screening, which, however, has high sensitivity but low specificity, resulting in a high rate of false-positive nodules. Combining PET with CT may provide more accurate diagnosis for indeterminate screening-detected nodules. In this work, we investigated low-dose dynamic 18F-FDG PET in discrimination between benign and malignant nodules using a virtual clinical trial based on patient study with ground truth. Six patients with initial LDCT screening-detected lung nodules received 90 min single-bed PET scans following a 10 mCi FDG injection. Low-dose static and dynamic images were generated from under-sampled list-mode data at various count levels (100%, 50%, 10%, 5%, and 1%). A virtual clinical trial was performed by adding nodule population variability, measurement noise, and static PET acquisition start time variability to the time activity curves (TACs) of the patient data. We used receiver operating characteristic (ROC) analysis to estimate the classification capability of standardized uptake value (SUV) and net uptake constant K i from their simulated benign and malignant distributions. Various scan durations and start times (t *) were investigated in dynamic Patlak analysis to optimize simplified acquisition protocols with a population-based input function (PBIF). The area under curve (AUC) of ROC analysis was higher with increased scan duration and earlier t *. Highly similar results were obtained using PBIF to those using image-derived input function (IDIF). The AUC value for K i using optimized t * and scan duration with 10% dose was higher than that for SUV with 100% dose. Our results suggest that dynamic PET with as little as 1 mCi FDG could provide discrimination between benign and malignant lung nodules with higher than 90% sensitivity and specificity for patients similar to the pilot and simulated population in this study, with LDCT screening-detected indeterminate lung nodules.