Idioma:

MoDL: Model-Based Deep Learning Architecture for Inverse Problems

Aggarwal, Hemant K. ; Mani, Merry P. ; Jacob, Mathews

IEEE transactions on medical imaging, 2019-02, Vol.38 (2), p.394-405

United States: IEEE

Texto completo disponible

Citas Citado por

Acciones

Título:
MoDL: Model-Based Deep Learning Architecture for Inverse Problems
Autor: Aggarwal, Hemant K. ; Mani, Merry P. ; Jacob, Mathews
Materias: Algorithms ; Animals ; Brain - diagnostic imaging ; Cats ; Conjugate gradients ; Consistency ; Convergence ; Convolution ; convolutional neural network ; Decoupling ; Deep Learning ; Dogs ; Humans ; Image processing ; Image Processing, Computer-Assisted - methods ; Image reconstruction ; Imaging ; Inverse problems ; Iterative methods ; Learning ; Machine learning ; Magnetic Resonance Imaging - methods ; Mathematical models ; Neural networks ; Numerical models ; Optimization ; parallel imaging ; Regularization ; Risk management ; Risk reduction ; Training ; Training data ; Weight
Es parte de: IEEE transactions on medical imaging, 2019-02, Vol.38 (2), p.394-405
Notas: ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
Descripción: 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 direct inversion approaches. Thus, reducing the demand for training data and training time. Since we rely on end-to-end training with weight sharing across iterations, the CNN weights are customized to the forward model, thus offering improved performance over approaches that rely on pre-trained denoisers. 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, primarily when the available GPU memory restricts the number of iterations.
Editor: United States: IEEE
Idioma: Inglés

Enlaces

View this record in MEDLINE/PubMed

Volver a la lista de resultados

Anterior Resultado 4 Siguiente Ir a la siguiente página

Realización: Logos de Redes Sociales:

MoDL: Model-Based Deep Learning Architecture for Inverse Problems

Aggarwal, Hemant K. ; Mani, Merry P. ; Jacob, Mathews

United States: IEEE

Buscando en bases de datos remotas, por favor espere