Recurrent Registration Neural Networks for Deformable Image Registration
Authors: Robin Sandkühler, Simon Andermatt, Grzegorz Bauman, Sylvia Nyilas, Christoph Jud, Philippe C. Cattin
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We trained our network on 2D magnetic resonance images of the lung and compared our method to a standard parametric B-spline registration. The experiments show, that our method performs on par for the accuracy but yields a more compact representation of the transformation. |
| Researcher Affiliation | Academia | Robin Sandkühler Department of Biomedical Engineering University of Basel, Switzerland Simon Andermatt Department of Biomedical Engineering University of Basel, Switzerland Grzegorz Bauman Division of Radiological Physics Department of Radiology University of Basel Hospital, Switzerland Sylvia Nyilas Pediatric Respiratory Medicine Department of Pediatrics Inselspital, Bern University Hospital University of Bern, Switzerland Christoph Jud Department of Biomedical Engineering University of Basel, Switzerland Philippe C. Cattin Department of Biomedical Engineering University of Basel, Switzerland |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described in this paper. While 'Airlab' is mentioned as a framework for comparison, no specific link or statement about the code for R2N2 is provided. |
| Open Datasets | No | The paper states 'Our data set consists of 48 lung acquisitions of 42 different patients.' and 'We used the data of 34 patients for the training set, 4 for the evaluation set, and 4 for the test set.', but it does not provide concrete access information (link, DOI, repository name, formal citation with authors/year) for this dataset. |
| Dataset Splits | Yes | We used the data of 34 patients for the training set, 4 for the evaluation set, and 4 for the test set. |
| Hardware Specification | Yes | The training of the network was performed on an NVIDIA Tesla V100 GPU. For the evaluation of the computation time for the registration of one image pair, we run both methods on an NVIDIA Ge Force GTX 1080. |
| Software Dependencies | No | The paper mentions 'Adam optimizer' and 'Air Lab framework' but does not provide specific version numbers for any software dependencies. |
| Experiment Setup | Yes | The network was trained in an unsupervised fashion for 180,000 iterations with a fixed sequence length of t = 25. We used the Adam optimizer [16] with the AMSGrad option [21] and a learning rate of 0.0001. The maximum shape size is set to σmax = 0.3 and the regularization weight to λR2N2 = 0.1. ... The B-spline registration use three spatial resolutions (64, 128, 256) with a kernel size of (7, 21, 57) pixels. As image loss the MSE and as regularizer the isotropic TV is used, with the regularization weight λBS = 0.01. We use the Adam optimizer [16] with the AMSGrad option [21], a learning rate of 0.001, and we perform 250 iterations per resolution level. |