MoDL: Model-Based Deep Learning Architecture for Inverse Problems
| Autor: | Hemant Kumar Aggarwal, Mathews Jacob, Merry Mani |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
FOS: Computer and information sciences
Network complexity Computer science Computer Vision and Pattern Recognition (cs.CV) Computer Science - Computer Vision and Pattern Recognition Overfitting Convolutional neural network Regularization (mathematics) Article 030218 nuclear medicine & medical imaging 03 medical and health sciences Deep Learning Dogs 0302 clinical medicine Image Processing Computer-Assisted Animals Humans Electrical and Electronic Engineering Training set Radiological and Ultrasound Technology business.industry Deep learning Brain Inverse problem Magnetic Resonance Imaging 3. Good health Computer Science Applications Cats Memory footprint Artificial intelligence business Algorithm Algorithms Software |
| Zdroj: | IEEE Transactions on Medical Imaging. 38:394-405 |
| ISSN: | 1558-254X 0278-0062 |
| Popis: | We introduce a model-based image reconstruction framework with a convolution neural network (CNN) based regularization prior. The proposed formulation provides a systematic approach for deriving deep architectures for inverse problems with the arbitrary structure. Since the forward model is explicitly accounted for, a smaller network with fewer parameters is sufficient to capture the image information compared to black-box deep learning approaches, thus reducing the demand for training data and training time. Since we rely on end-to-end training, the CNN weights are customized to the forward model, thus offering improved performance over approaches that rely on pre-trained denoisers. The main difference of the framework from existing end-to-end training strategies is the sharing of the network weights across iterations and channels. Our experiments show that the decoupling of the number of iterations from the network complexity offered by this approach provides benefits including lower demand for training data, reduced risk of overfitting, and implementations with significantly reduced memory footprint. We propose to enforce data-consistency by using numerical optimization blocks such as conjugate gradients algorithm within the network; this approach offers faster convergence per iteration, compared to methods that rely on proximal gradients steps to enforce data consistency. Our experiments show that the faster convergence translates to improved performance, especially when the available GPU memory restricts the number of iterations. Comment: published in IEEE Transaction on Medical Imaging |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|