| dc.description.abstract | Attenuation correction is a critical phenomenon in quantitative PET imaging with its own special challenges. However, computerized tomography (CT) modality which is used for attenuation correction and anatomical localization increases patient radiation dose. This study was aimed to develop a deep learning model for attenuation correction of whole-body 68Ga-DOTATATE PET images.
Methods: Non-attenuation corrected (NAC) and CT-based attenuation corrected (CTAC) whole-body 68Ga-DOTATATE PET images of 118 patients from two different imaging centers were used. We implemented a residual deep learning model using the NiftyNet framework. The model was trained four times and evaluated six times using the test data from the centers. Quality of the synthesized PET images were compared with the PET-CTAC images using different evaluation metrics, including peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), mean square error (MSE), and root mean square error (RMSE).
Results: Quantitative analysis of four network training sessions and six evaluations revealed the highest and lowest PSNR values as (52.86 ± 6.6) and (47.96 ± 5.09), respectively. Similarly, the highest and lowest SSIM values were obtained (0.99 ± 0.003) and (0.97 ± 0.01), respectively. Additionally, the highest and lowest RMSE and MSE values fell within the ranges of (0.0117 ± 0.003), (0.0015 ± 0.000103), and (0.01072 ± 0.002), (0.000121 ± 5.07×e^(-5)), respectively. The results indicated that the highest PSNR was associated with the training and testing datasets originating from the same center. Conversely, the evaluations that involved training and testing datasets from different centers exhibited the lowest PSNR and SSIM values, while demonstrating the highest MSE and RMSE values. Furthermore, the evaluations
with the highest SSIM and the lowest MSE and RMSE were associated with scenarios in which datasets from both centers were used for network training and testing. | en_US |
