Direct inference of Patlak parametric images in whole-body PET/CT imaging using convolutional neural networks

Abstract

Purpose This study proposed and investigated the feasibility of estimating Patlak-derived influx rate constant (Ki) from standardized uptake value (SUV) and/or dynamic PET image series. Methods Whole-body 18F-FDG dynamic PET images of 19 subjects consisting of 13 frames or passes were employed for training a residual deep learning model with SUV and/or dynamic series as input and Ki-Patlak (slope) images as output. The training and evaluation were performed using a nine-fold cross-validation scheme. Owing to the availability of SUV images acquired 60 min post-injection (20 min total acquisition time), the data sets used for the training of the models were split into two groups: “With SUV” and “Without SUV.” For “With SUV” group, the model was first trained using only SUV images and then the passes (starting from pass 13, the last pass, to pass 9) were added to the training of the model (one pass each time). For this group, 6 models were developed with input data consisting of SUV, SUV plus pass 13, SUV plus passes 13 and 12, SUV plus passes 13 to 11, SUV plus passes 13 to 10, and SUV plus passes 13 to 9. For the “Without SUV” group, the same trend was followed, but without using the SUV images (5 models were developed with input data of passes 13 to 9). For model performance evaluation, the mean absolute error (MAE), mean error (ME), mean relative absolute error (MRAE%), relative error (RE%), mean squared error (MSE), root mean squared error (RMSE), peak signal-to-noise ratio (PSNR), and structural similarity index (SSIM) were calculated between the predicted Ki-Patlak images by the two groups and the reference Ki-Patlak images generated through Patlak analysis using the whole acquired data sets. For specific evaluation of the method, regions of interest (ROIs) were drawn on representative organs, including the lung, liver, brain, and heart and around the identified malignant lesions. Results The MRAE%, RE%, PSNR, and SSIM indices across all patients were estimated as 7.45 ± 0.94%, 4.54 ± 2.93%, 46.89 ± 2.93, and 1.00 ± 6.7 × 10−7, respectively, for models predicted using SUV plus passes 13 to 9 as input. The predicted parameters using passes 13 to 11 as input exhibited almost similar results compared to the predicted models using SUV plus passes 13 to 9 as input. Yet, the bias was continuously reduced by adding passes until pass 11, after which the magnitude of error reduction was negligible. Hence, the predicted model with SUV plus passes 13 to 9 had the lowest quantification bias. Lesions invisible in one or both of SUV and Ki-Patlak images appeared similarly through visual inspection in the predicted images with tolerable bias. Conclusion This study concluded the feasibility of direct deep learning-based approach to estimate Ki-Patlak parametric maps without requiring the input function and with a fewer number of passes. This would lead to shorter acquisition times for WB dynamic imaging with acceptable bias and comparable lesion detectability performance.

A deep neural network for parametric image reconstruction on a large axial field-of-view PET

“…In this study, present results have proven that utilizing sinogram and dynamic images simultaneously could deliver high-quality parametric images for the DeepPET-like network. In addition, we explored the feasibility of CNN-based parametric image generation from static or dynamic PET images only [53][54]. A 2D U-Net CNN [55] Around the deep learning based parametric imaging researches, embedding a CNN module into reconstruction model, like CTguided Logan plot [56], in which an iterative reconstruction framework with a deep neural network as a constraint was implemented.…”

Quantitation of Dynamic Total-Body PET Imaging: Recent Developments and Future Perspectives

“…Another research interest in future work is the implementation of artificial intelligence (AI) for the totalbody PET imaging [87,150]. As a subcategory of AI, deep learning (DL) techniques, e.g., convolutional neural network (CNN) [151] and generative adversarial network (GAN) [89], have been extensively used in PET for solving a wide variety of problems involving image reconstruction [152][153][154], denoising [155,156], segmentation [157,158] as well as quantitation [159,160]. A few initial attempts have been made to extract the flux (K i ) from total-body PET studies by DL methods [61,148,161].…”

Kinetic parameter estimation of hepatocellular carcinoma on 18F‐FDG PET/CT based on Bayesian method

“…Blood glucose levels were verified after the patients had fasted for at least 6 h. A 5-minute dynamic PET/CT scan was performed before conventional PET/CT imaging for all patients. A bolus injection of 18 F-FDG (5.5 MBq/kg, estimated radiation dose = 7.3 mSv 26 ) in 2 ml of 0.9% saline was performed and flushed with 20 ml of 0.9% saline at a flow rate of 2 ml/s. A liver CT scan (120 kV, 100 mA, dose length product (DLP) = 159.8 mGy⋅cm, estimated radiation dose = 2.4 mSv 27 ) was performed in a single bed with the liver at the center of the scanner's field of view.…”

“…It should be noted that in the PET literature, K 1 is often reported using a capital K to denote a different unit of measurement. [28][29][30] k 2 (L/min) is the clearance rate of 18 F-FDG returning from the liver tissue to the blood. k 3 (L/min) represents the phosphorylation rate of 18 F-FDG into 18 F-FDG 6-phosphate by hexokinase,and k 4 (L/min) is the dephosphorylation rate by phosphatase.…”

“…[28][29][30] k 2 (L/min) is the clearance rate of 18 F-FDG returning from the liver tissue to the blood. k 3 (L/min) represents the phosphorylation rate of 18 F-FDG into 18 F-FDG 6-phosphate by hexokinase,and k 4 (L/min) is the dephosphorylation rate by phosphatase. Here, the kinetic parameters K 1 −k 4 and the hepatic artery blood supply fraction fa, where V b is the fraction of the measured volume occupied by blood, are the parameters to be determined.…”