TY - JOUR T1 - Denoising of Scintillation Camera Images using a Deep Convolutional Neural Network: A Monte Carlo Simulation Approach JF - Journal of Nuclear Medicine JO - J Nucl Med DO - 10.2967/jnumed.119.226613 SP - jnumed.119.226613 AU - David Minarik AU - Olof Enqvist AU - Elin Trägårdh Y1 - 2019/07/01 UR - http://jnm.snmjournals.org/content/early/2019/07/18/jnumed.119.226613.abstract N2 - Rationale: Scintillation camera images contain a large amount of Poisson noise. We have investigated whether noise can be removed in whole-body bone scans using convolutional neural networks (CNN) trained with sets of noisy and noiseless images obtained by Monte Carlo simulation. Methods: Three CNNs were generated using three different sets of training images: simulated bone scan images, images of a cylindrical phantom with hot and cold spots, and a mix of the first two. Each training set consisted of 40,000 noiseless and noisy image pairs. The CNNs were evaluated with simulated images of a cylindrical phantom and simulated bone scan images. The mean squared error (MSE) between filtered and true images was used as difference metric and the coefficient of variation (COV) was used to estimate noise reduction. The CNNs were compared to Gaussian and median filters. A clinical evaluation was performed where the ability to detect metastases for CNN- and Gaussian-filtered bone scans with half the number of counts were compared with standard bone scans. Results: The best CNN reduced COV with on average 92%, and the best standard filter reduced COV with 88%. The best CNN gave an MSE that was on average 68% and 20% better than the best standard filters, for the cylindrical and bone scan images, respectively. The best CNN for the cylindrical phantom and bone scans were the dedicated CNNs. No significant differences for the ability to detect metastases between standard, CNN- and Gaussian-filtered bone scans were found. Conclusion: Noise can be removed efficiently regardless of noise level with little or no resolution loss. The CNN filter enables reducing the scanning time by half and still obtain good accuracy for bone metastases assessment. ER -