In Physics in medicine and biology
Our study aims to improve the signal-to-noise ratio (SNR) of PET imaging using conditional unsupervised learning. The proposed method does not require low- and high-quality pairs for network training which can be easily applied to existing PET/CT and PET/MR datasets. This method includes two steps: populational training and individual fine-tuning. As for populational training, a network was first pre-trained by a group of patients' noisy PET images and the corresponding anatomical prior images from CT or MR. As for individual fine-tuning, a new network with initial parameters inherited from the pre-trained network was fine-tuned by the test patient's noisy PET image and the corresponding anatomical prior image. Only the last few layers were fine-tuned to take advantage of the populational information and the pre-training efforts. Both networks took the CT or MR images as the network input so that the network output was conditioned on the patient's anatomic prior information. The noisy PET images were used as the training and fine-tuning labels. The proposed method was evaluated on a 68Ga-PRGD2 PET/CT dataset and a 18F-FDG PET/MR dataset. For the PET/CT dataset, with the original noisy PET image as the baseline, the proposed method has a significantly higher contrast to noise ratio (CNR) improvement (64.45% ± 21.98%) than Gaussian (12.64% ± 6.15%, P = 0.002), nonlocal mean method (24.35% ± 16.30%, P = 0.002) and conditional deep image prior method (53.35% ± 21.78%, P = 0.037). For the PET/MR dataset, compared to Gaussian (18.16% ± 10.02%, P < 0.0001), NLM (25.36% ± 19.48%, P < 0.0001) and CDIP (46.80% ± 25.23%, P < 0.0001), the CNR improvement ratio of the proposed method (57.79% ± 28.28%) is the highest. In addition, the denoised images also showed that the proposed method can accurately restore tumor structures while also smoothing out the noise.
Cui Jianan, Gong Kuang, Guo Ning, Wu Chenxi, Kim Kyungsang, Liu Huafeng, Li Quanzheng
Deep neural network, Denoising, PET and anatomical pair, Positron Emission Tomography, Unsupervised deep learning