PET Denoising and Uncertainty Estimation Based on NVAE Model Using Quantile Regression Loss
Jianan Cui,Yutong Xie,Anand A. Joshi,Kuang Gong,Kyungsang Kim,Young-Don Son,Jong-Hoon Kim,Richard Leahy,Huafeng Liu,Quanzheng Li
DOI: https://doi.org/10.1007/978-3-031-16440-8_17
2022-01-01
Abstract:Deep learning-based methods have shown their superior performance for medical imaging, but their clinical application is still rare. One reason may come from their uncertainty. As data-driven models, deep learning-based methods are sensitive to imperfect data. Thus, it is important to quantify the uncertainty, especially for positron emission tomography (PET) denoising tasks where the noise is very similar to small tumors. In this paper, we proposed a Nouveau variational autoencoder (NVAE) based model using quantile regression loss for simultaneous PET image denoising and uncertainty estimation. Quantile regression loss was performed as the reconstruction loss to avoid the variance shrinkage problem caused by the traditional reconstruction probability loss. The variance and mean can be directly calculated from the estimated quantiles under the Logistic assumption, which is more efficient than Monte Carlo sampling. Experiment based on real 11 C-DASB datasets verified that the denoised PET images of the proposed method have a higher mean(+/- SD) peak signal-to-noise ratio (PSNR) (40.64 +/- 5.71) and structural similarity index measure (SSIM) (0.9807 +/- 0.0063) than Unet-based denoising (PSNR, 36.18 +/- 5.55; SSIM, 0.9614 +/- 0.0121) and NVAE model using Monte Carlo sampling (PSNR, 37.00 +/- 5.35; SSIM, 0.9671 +/- 0.0095) methods.