Machine learning and deep learning for image reconstruction: PART 1 (convolutional neural networks)

Review of conventional reconstruction and its limitations Machine learning principles Linear operator example: matrix, and convolution mappings Including non-linearities - convolutional neural networks (CNNs) Conventional reconstruction: PET: suffers from noise due to low dose or short scan time frames MRI: suffers from aliasing artefacts from undersampling in faster scanning Resolved by regularisation methods – but how can these be optimised? Machine learning: Basic intro to principles of learning entire