Deep learning-based PET scatter estimation systems and methods using an inception neural network model

Systems, apparatus, methods, and computer-readable storage media to estimate scatter in an image are disclosed. An example apparatus includes a network generator to generate and train an inception neural network using first and second input to deploy an inception neural network model to process image data when first and second outputs of the inception neural network converge, the first input based on a raw sinogram of a first image and the second input based on an attenuation-corrected sinogram of the first image, the inception neural network including a first filter of a first size and a second filter of a second size in a layer to process the first input and/or the second input to generate an estimate of scatter in the first image. The example apparatus also includes an image processor to apply the estimate of scatter to a second image to generate a processed image.