NiftyNet: a deep-learning platform for medical imaging

Gibson, Eli; Li, Wenqi; Sudre, Carole H.; Fidon, Lucas; Shakir, Dzhoshkun I.; Wang, Guotai; Eaton-Rosen, Zach; Gray, Robert; Doel, Tom; Hu, Yipeng; Whyntie, T.; Nachev, Parashkev; Modat, Marc; Barratt, Dean C.; Ourselin, Sébastien; Cardoso, M. Jorge; Vercauteren, Tom

doi:10.1016/j.cmpb.2018.01.025

Cited by 509 publications

(315 citation statements)

References 49 publications

(65 reference statements)

Supporting

Mentioning

314

Contrasting

Unclassified

Order By: Relevance

“…Geometric augmentations were sampled independently and combined as one affine transform, using random rotations (all axis ranging from -10 to 10 degrees), random shears ([0.5, 0.5]) and random scaling ([0.75, 1.5]). For the non-geometric augmentations we applied k-space motion artefact augmentation as described in [12] and bias field augmentation as implemented in [3]. We measure how useful these additional augmentations are in our experiments.…”

Section: Methodsmentioning

confidence: 99%

Multi-domain Adaptation in Brain MRI Through Paired Consistency and Adversarial Learning

Orbes-Arteaga

Varsavsky

Sudre

et al. 2019

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Supervised learning algorithms trained on medical images will often fail to generalize across changes in acquisition parameters. Recent work in domain adaptation addresses this challenge and successfully leverages labeled data in a source domain to perform well on an unlabeled target domain. Inspired by recent work in semi-supervised learning we introduce a novel method to adapt from one source domain to n target domains (as long as there is paired data covering all domains). Our multi-domain adaptation method utilises a consistency loss combined with adversarial learning. We provide results on white matter lesion hyperintensity segmentation from brain MRIs using the MICCAI 2017 challenge data as the source domain and two target domains. The proposed method significantly outperforms other domain adaptation baselines.

show abstract

Section: Methodsmentioning

confidence: 99%

Multi-domain Adaptation in Brain MRI Through Paired Consistency and Adversarial Learning

Orbes-Arteaga

Varsavsky

Sudre

et al. 2019

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

show abstract

“…We used Tensorflow and implemented our models within the NiftyNet framework [41]. Models were trained on NVIDIA Titan Xp, P6000 and V100.…”

Section: A4 Implementation Detailsmentioning

confidence: 99%

Stochastic Filter Groups for Multi-Task CNNs: Learning Specialist and Generalist Convolution Kernels

Bragman

Tanno

Ourselin

et al. 2019

2019 IEEE/CVF International Conference on Computer Vision (ICCV)

View full text Add to dashboard Cite

The performance of multi-task learning in Convolutional Neural Networks (CNNs) hinges on the design of feature sharing between tasks within the architecture. The number of possible sharing patterns are combinatorial in the depth of the network and the number of tasks, and thus hand-crafting an architecture, purely based on the human intuitions of task relationships can be time-consuming and suboptimal. In this paper, we present a probabilistic approach to learning task-specific and shared representations in CNNs for multi-task learning. Specifically, we propose "stochastic filter groups" (SFG), a mechanism to assign convolution kernels in each layer to "specialist" or "generalist" groups, which are specific to or shared across different tasks, respectively. The SFG modules determine the connectivity between layers and the structures of task-specific and shared representations in the network. We employ variational inference to learn the posterior distribution over the possible grouping of kernels and network parameters. Experiments demonstrate that the proposed method generalises across multiple tasks and shows improved performance over baseline methods.

show abstract

“…Subjects were randomly split between training (2420 simulated volumes over 19 subjects), validation (242 simulated volumes over two subjects) and testing (726 simulated volumes over six subjects). Networks were trained until convergence, as defined per performance on the validation set when over 1000 iterations have elapsed without decreases in the loss (a probabilistic version of the dice loss [11]), using NiftyNet [7], a deep learning framework designed for medical imaging. In a first instance, we compare the performance of 3D U-Net-Physics with that of base 3D U-Net (trained simply by excluding the physics branch), as well as GIF [2], a segmentation software based on geodesical information flows.…”

Section: Robustness To Acquisition Parametersmentioning

confidence: 99%

Physics-Informed Brain MRI Segmentation

Borges¹,

Sudre²,

Varsavsky³

et al. 2019

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Magnetic Resonance Imaging (MRI) is one of the most flexible and powerful medical imaging modalities. This flexibility does however come at a cost; MRI images acquired at different sites and with different parameters exhibit significant differences in contrast and tissue appearance, resulting in downstream issues when quantifying brain anatomy or the presence of pathology. In this work, we propose to combine multiparametric MRI-based static-equation sequence simulations with segmentation convolutional neural networks (CNN), to make these networks robust to variations in acquisition parameters. Results demonstrate that, when given both the image and their associated physics acquisition parameters, CNNs can produce segmentations that exhibit robustness to acquisition variations. We also show that the proposed physics-informed methods can be used to bridge multi-centre and longitudinal imaging studies where imaging acquisition varies across a site or in time.

show abstract

NiftyNet: a deep-learning platform for medical imaging

Cited by 509 publications

References 49 publications

Multi-domain Adaptation in Brain MRI Through Paired Consistency and Adversarial Learning

Multi-domain Adaptation in Brain MRI Through Paired Consistency and Adversarial Learning

Stochastic Filter Groups for Multi-Task CNNs: Learning Specialist and Generalist Convolution Kernels

Physics-Informed Brain MRI Segmentation

Contact Info

Product

Resources

About