A Deep Learning Approach for Noise Suppression in Cardiac-gated SPECT Studies

This study investigates the effectiveness of a deep learning (DL) approach for noise suppression in cardiac-gated SPECT imaging, aiming to enhance image quality without distorting left ventricular (LV) functional measurements. The authors adapted a 3D convolutional autoencoder network, demonstrating significant improvements in image uniformity and reduced temporal variability, while maintaining high correlation in LV function metrics before and after denoising. The findings suggest that this DL method is applicable across different imaging systems, offering a promising solution for noise reduction in clinical cardiac studies.