NimbleReg: A light-weight deep-learning framework for diffeomorphic image registration

Antoine Legouhy; Hojjat Azadbakht; Hui Zhang; Nolah Mazet; Ross Callaghan; Vivien Julienne

arxiv: 2503.07768 · v2 · submitted 2025-03-10 · 💻 cs.CV

NimbleReg: A light-weight deep-learning framework for diffeomorphic image registration

Antoine Legouhy , Ross Callaghan , Nolah Mazet , Vivien Julienne , Hojjat Azadbakht , Hui Zhang This is my paper

Pith reviewed 2026-05-22 23:58 UTC · model grok-4.3

classification 💻 cs.CV

keywords diffeomorphic registrationdeep learningsurface representationPointNetmedical image registrationstationary velocity fieldlight-weight model

0 comments

The pith

NimbleReg registers images diffeomorphically from surface points of multiple segmented regions using a light-weight PointNet model.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents a deep learning approach to image registration that works from boundary surfaces rather than full image grids. It takes surfaces extracted from several segmented anatomical regions and passes them through a PointNet network to produce a stationary velocity field. That field defines a single diffeomorphic transformation that applies across the whole space. The resulting model is much lighter than typical grid-based networks yet produces alignment quality comparable to state-of-the-art image-consuming methods.

Core claim

NimbleReg shows that surfaces from multiple segmented regions can be fused by a PointNet backbone into one stationary velocity field that generates a diffeomorphic transformation defined over the entire ambient space, achieving alignment comparable to state-of-the-art DL-based registration techniques that consume images.

What carries the argument

PointNet backbone combined with stationary velocity field parametrization to fuse multiple regional surface mappings into one diffeomorphic transformation over the whole space.

If this is right

Registration becomes feasible on hardware with limited memory because full image grids are not required.
Multiple segmented regions can be aligned simultaneously under one consistent transformation.
Diffeomorphic properties are guaranteed by the stationary velocity field parametrization.
The method can exploit the low-cost fine-grained segmentations now widely available.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The surface-only input could make the method more robust to intensity inhomogeneities that affect image-based networks.
The reduced compute footprint may enable registration inside time-critical clinical workflows.
Similar surface-to-velocity-field pipelines could be tested on non-medical point-cloud alignment tasks.

Load-bearing premise

Surface representations from multiple segmented regions can be fused by the PointNet-plus-stationary-velocity-field pipeline into a single diffeomorphic transformation defined over the entire ambient space.

What would settle it

A test showing that the generated transformation either fails to align points outside the input surfaces or violates topology preservation in the ambient space.

Figures

Figures reproduced from arXiv: 2503.07768 by Antoine Legouhy, Hojjat Azadbakht, Hui Zhang, Nolah Mazet, Ross Callaghan, Vivien Julienne.

**Figure 1.** Figure 1: a). 2. The individual region surfaces are then merged into a single mesh, stitched together using the common interface points. 3. The overall mesh is smoothed. This approach, rather than smoothing the individual surfaces independently, ensures that the interfaces between adjacent regions are preserved. 4. A pre-alignment onto a standard reference is performed using Polaffini [14], which estimates an initi… view at source ↗

**Figure 1.** Figure 1: a) Synthetic segmentation and associated extracted surfaces. b) Velocities estimated by fθ for each region and the associated SVF; c) Deformed points and grid after integration. d) Deformed points and grid without integration. 2.4 Loss functions To define a distance metric between point clouds without one-to-one correspondences, a common heuristic from the Iterative Closest Point (ICP) algorithm [18] is … view at source ↗

**Figure 2.** Figure 2: Diagrams for the proposed registration model at training (bottom-left) and at inference (right), and a zoom in on fθ’s architecture (top-left). SynthMorph [4]. We also included in the comparison two initializations based on region centroids using Polaffini [14]: Affine (σ = ∞, similar to [5]) and Polyaffine (σ following Silverman’s rule of thumb). We evaluated the quality of the alignment using two metrics… view at source ↗

**Figure 3.** Figure 3: Quality of alignment metrics between moved and reference. Left: segmentation overlap (higher the better). Right: surface distance (lower the better). Lines are colored according to which datasets the images are from. which necessitates the computation of pairwise distances. However, this burden is alleviated at inference using KD-trees. 4 Discussion and conclusion We presented NimbleReg, a novel deep-learn… view at source ↗

read the original abstract

This paper presents NimbleReg, a light-weight deep-learning (DL) framework for diffeomorphic image registration leveraging surface representation of multiple segmented anatomical regions. Deep learning has revolutionized image registration but most methods typically rely on cumbersome gridded representations, leading to hardware-intensive models. Reliable fine-grained segmentations, that are now accessible at low cost, are often used to guide the alignment. Light-weight methods representing segmentations in terms of boundary surfaces have been proposed, but they lack mechanism to support the fusion of multiple regional mappings into an overall diffeomorphic transformation. Building on these advances, we propose a DL registration method capable of aligning surfaces from multiple segmented regions to generate an overall diffeomorphic transformation for the whole ambient space. The proposed model is light-weight thanks to a PointNet backbone. Diffeomoprhic properties are guaranteed by taking advantage of the stationary velocity field parametrization of diffeomorphisms. We demonstrate that this approach achieves alignment comparable to state-of-the-art DL-based registration techniques that consume images.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

NimbleReg sketches a PointNet model that fuses multi-region surfaces into one SVF for diffeomorphic registration, but the abstract supplies no metrics and the fusion step is not described.

read the letter

NimbleReg proposes a PointNet-driven approach to diffeomorphic registration that works from surface points of multiple anatomical regions instead of full image grids. The goal is a lighter model that still produces a guaranteed diffeomorphic transform via stationary velocity fields. What stands out as new is the attempt to fuse mappings from several regions through a shared PointNet backbone into one overall SVF. Prior surface-based methods apparently didn't handle multiple regions this way. The paper does a decent job framing the problem: DL registration is heavy on compute, segmentations are cheap now, and SVF is a solid way to enforce diffeomorphism. That part is clear. The soft spots are bigger. The abstract claims performance comparable to image-based SOTA but gives no numbers at all—no errors, no stats, no datasets. That's not enough to evaluate the claim. More importantly, the fusion mechanism is not described. How do the per-region PointNet features combine into a single velocity field on the grid? Is it predicted densely or interpolated from surfaces? Without that, it's hard to see why the result is diffeomorphic everywhere, not just near the surfaces. The stress-test concern holds up based on what's here. This paper is for researchers in medical imaging who want lighter registration tools when they have segmentations. A reader focused on practical DL methods might find the architecture worth looking at if the full paper has the missing details. It deserves a serious referee because the idea is coherent on its face and builds on real techniques, even if the current write-up is thin on evidence. I would recommend sending it to peer review so the authors can address the fusion details and add quantitative results.

Referee Report

2 major / 1 minor

Summary. The paper presents NimbleReg, a lightweight DL framework for diffeomorphic image registration that processes surface point clouds from multiple segmented anatomical regions using a shared PointNet backbone to produce a stationary velocity field (SVF) whose exponential map yields a diffeomorphism over the full ambient image space. It claims this surface-based approach achieves alignment performance comparable to state-of-the-art image-based DL registration methods while being computationally lighter.

Significance. If the fusion mechanism and performance claims are substantiated with quantitative evidence, the work could offer a meaningful efficiency gain for diffeomorphic registration by shifting from dense image grids to sparse surface representations, with the SVF parametrization providing a standard route to diffeomorphism guarantees. The multi-region fusion via PointNet is a potentially useful extension of prior surface-based methods, but its load-bearing details remain undemonstrated.

major comments (2)

[Abstract and method description] Abstract and method description (paragraph on fusion mechanism): the central claim that multiple regional surface mappings are fused into one global diffeomorphism defined over the entire ambient space lacks any explicit construction for combining PointNet outputs into a single dense SVF (e.g., whether features are concatenated into a shared decoder, how the velocity field is obtained on the image grid versus interpolated from surfaces, or what regularization ensures positive Jacobian and invertibility everywhere). This directly undermines both the diffeomorphism guarantee and the comparability to image-based methods.
[Abstract] Abstract: the assertion that the approach 'achieves alignment comparable to state-of-the-art DL-based registration techniques that consume images' is presented without any quantitative metrics, error bars, datasets, ablation studies, or baseline comparisons, leaving the primary empirical claim unsupported and unverifiable from the provided text.

minor comments (1)

[Abstract] Abstract contains a typo: 'Diffeomoprhic' should be 'Diffeomorphic'.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight areas where the manuscript can be clarified. We respond to each major comment below and will make corresponding revisions.

read point-by-point responses

Referee: [Abstract and method description] Abstract and method description (paragraph on fusion mechanism): the central claim that multiple regional surface mappings are fused into one global diffeomorphism defined over the entire ambient space lacks any explicit construction for combining PointNet outputs into a single dense SVF (e.g., whether features are concatenated into a shared decoder, how the velocity field is obtained on the image grid versus interpolated from surfaces, or what regularization ensures positive Jacobian and invertibility everywhere). This directly undermines both the diffeomorphism guarantee and the comparability to image-based methods.

Authors: We agree that the fusion mechanism requires a more explicit description to support the diffeomorphism claim. The manuscript uses a shared PointNet to process multi-region point clouds and predict SVF parameters, with the exponential map providing the diffeomorphism. However, the details of densifying the velocity field to the full image grid and the regularization for ensuring positive Jacobians are not fully elaborated. In the revised manuscript we will add a dedicated paragraph in the Methods section describing the decoder architecture, grid interpolation step, and regularization terms. revision: yes
Referee: [Abstract] Abstract: the assertion that the approach 'achieves alignment comparable to state-of-the-art DL-based registration techniques that consume images' is presented without any quantitative metrics, error bars, datasets, ablation studies, or baseline comparisons, leaving the primary empirical claim unsupported and unverifiable from the provided text.

Authors: The Experiments section of the manuscript contains the supporting quantitative results, including Dice scores, target registration errors with error bars, dataset descriptions, and comparisons to image-based baselines. To make the abstract self-contained and address the concern, we will revise it to include a concise statement of the key metrics and evaluation setup. revision: yes

Circularity Check

0 steps flagged

No circularity; method architecture and empirical claims are independent of reported outcomes

full rationale

The paper describes a PointNet-based model with stationary velocity field parametrization to produce diffeomorphic transformations from fused surface representations of segmented regions. No equations, fitted parameters, or self-citations are shown to reduce the claimed alignment performance or diffeomorphism guarantee to the input data by construction. The fusion step and SVF choice are presented as design decisions whose validity is assessed via external comparison to image-based SOTA methods, making the derivation self-contained without load-bearing circular steps.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claim rests on the standard mathematical properties of stationary velocity fields for diffeomorphisms and on the empirical assumption that PointNet can learn a useful mapping from surface points; no new entities are postulated and the only free parameters are the usual network weights trained on data.

free parameters (1)

network weights
Standard trainable parameters of the PointNet backbone fitted during supervised or unsupervised training on registration pairs.

axioms (2)

standard math Stationary velocity field integration yields a diffeomorphism
Invoked in the abstract when stating that diffeomorphic properties are guaranteed by the SVF parametrization.
domain assumption Surface points from multiple regions contain sufficient information to drive whole-volume alignment
Central modeling choice stated in the abstract when moving from image grids to boundary surfaces.

pith-pipeline@v0.9.0 · 5722 in / 1346 out tokens · 34572 ms · 2026-05-22T23:58:17.706098+00:00 · methodology

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

SpaCeFormer: Fast Proposal-Free Open-Vocabulary 3D Instance Segmentation
cs.CV 2026-04 unverdicted novelty 6.0

SpaCeFormer delivers 11.1 zero-shot mAP on ScanNet200 (2.8x prior proposal-free best) and runs 2-3 orders of magnitude faster than multi-stage 2D+3D pipelines by using spatial window attention and Morton-curve seriali...

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages · cited by 1 Pith paper

[1]

D., Berendsen, F

De Vos, B. D., Berendsen, F. F., Viergever, M. A., Staring, M., Išgum, I. End- to-End Unsupervised Deformable Image Registration with a Convolutional Neural Network. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support LNCS, vol 10553. (2017)

work page 2017
[2]

D., Berendsen, F

De Vos, B. D., Berendsen, F. F., Viergever, M. A., Sokooti, H., Staring, M., Išgum, I. A deep learning framework for unsupervised affine and deformable image regis- tration. Medical image analysis, 52, 128-143 (2019)

work page 2019
[3]

Weakly-supervised convolutional neural networks for multimodal image registra- tion

Hu, Y., Modat, M., Gibson, E., Li, W., Ghavami, N., Bonmati, E., ..., Vercauteren, T. Weakly-supervised convolutional neural networks for multimodal image registra- tion. Medical image analysis, 49, 1-13 (2018)

work page 2018
[4]

N., Iglesias, J

Hoffmann, M., Billot, B., Greve, D. N., Iglesias, J. E., Fischl, B., Dalca, A. V. Syn- thMorph: learning contrast-invariant registration without acquired images. Trans- actions on Medical Imaging, 41(3), 543-558 (2021)

work page 2021
[5]

A ready-to-use machine learning tool for symmetric multi-modality registration of brain MRI

Iglesias, J.E. A ready-to-use machine learning tool for symmetric multi-modality registration of brain MRI. Scientific Reports, 13(1), 6657 (2023)

work page 2023
[6]

M., Hu, Y

Baum, Z. M., Hu, Y. and Barratt, D. C. Multimodality biomedical image registra- tion using free point transformer networks. In Medical Ultrasound, and Preterm, Perinatal and Paediatric Image Analysis. First International Workshop, ASMUS 2020, PIPPI 2020, 116-125 (2020)

work page 2020
[7]

M., Saeed, S

Min, Z., Baum, Z. M., Saeed, S. U., Emberton, M., Barratt, D. C., Taylor, Z. A. and Hu, Y. Biomechanics-informed Non-rigid Medical Image Registration and its Inverse Material Property Estimation with Linear and Nonlinear Elasticity. MICCAI 2024, 564-574 (2024)

work page 2024
[8]

A Log-Euclidean Framework for Statistics on Diffeomorphisms

Arsigny, V., Commowick, O., Pennec, X., Ayache, N. A Log-Euclidean Framework for Statistics on Diffeomorphisms. MICCAI 2006, LNCS vol 4190 (2006)

work page 2006
[9]

Contributions to 3D diffeomorphic atlas esti- mation: application to brain images

Bossa, M., Hernandez, M., Olmos, S. Contributions to 3D diffeomorphic atlas esti- mation: application to brain images. MICCAI, Proceedings, Part I 10, pp. 667-674 (2007)

work page 2007
[10]

Symmetric log-domain dif- feomorphic registration: A demons-based approach

Vercauteren, T., Pennec, X., Perchant, A., Ayache, N. Symmetric log-domain dif- feomorphic registration: A demons-based approach. MICCAI, pp. 754-761 (2008)

work page 2008
[11]

Unsupervised learning of probabilistic diffeomorphic registration for images and surfaces, Medical Image Analysis, vol 57, 226-236 (2019)

Dalca, A.V., Balakrishnan, G., Guttag, J., Sabuncu, M.R. Unsupervised learning of probabilistic diffeomorphic registration for images and surfaces, Medical Image Analysis, vol 57, 226-236 (2019)

work page 2019
[12]

Yang, X., Li, Y., Reutens, D. et al. Diffeomorphic Metric Landmark Mapping Using Stationary Velocity Field Parameterization. Int J Comput Vis 115, 69–86 (2015)

work page 2015
[13]

Lorensen, W. E. and Cline H. E. Marching cubes: A high resolution 3D surface construction algorithm. SIGGRAPH Comput. Graph. 21, 4, 163–169 (1987)

work page 1987
[14]

POLAFFINI: Efficient Feature-Based Polyaffine Initialization for Improved Non-linear Image Registration

Legouhy, A., Callaghan, R., Azadbakht, H., Zhang, H. POLAFFINI: Efficient Feature-Based Polyaffine Initialization for Improved Non-linear Image Registration. IPMI 2023. LNCS, vol 13939 (2023)

work page 2023
[15]

Arsigny, V., Commowick, O., Ayache, N. et al. A Fast and Log-Euclidean Polyaffine Framework for Locally Linear Registration. J Math Imaging Vis 33, 222–238 (2009). 10 A. Legouhy et al

work page 2009
[16]

R., Su, H., Mo, K., Guibas, L.J

Qi, C. R., Su, H., Mo, K., Guibas, L.J. PointNet: deep learning on point sets for 3D classification and segmentation. CVPR, 652–660 (2017)

work page 2017
[17]

A robust and efficient block matching framework for non linear registration of thoracic CT images

Garcia, V., Commowick, O., Malandain, G. A robust and efficient block matching framework for non linear registration of thoracic CT images. In Grand Challenges in Medical Image Analysis (MICCAI workshop) pp. 1-10 (2010)

work page 2010
[18]

Besl, P. J. and McKay, N. D. Method for registration of 3-D shapes. Sensor fusion IV: control paradigms and data structures, vol. 1611, pp. 586-606 (1992)

work page 1992
[19]

Alzheimer’s Disease Neuroimaging Initiative (ADNI): clinical char- acterization

Petersen, R.C., Aisen, P.S., Beckett, L.A., Donohue, M.C., Gamst, A.C., Harvey, D.J., Jack, C.R., Jr, Jagust, W.J., Shaw, L.M., Toga, A.W., Trojanowski, J.Q., Weiner, M.W. Alzheimer’s Disease Neuroimaging Initiative (ADNI): clinical char- acterization. Neurology, 74(3), 201–209 (2010)

work page 2010
[20]

101 Labeled Brain Images and a Consistent Human Cortical Labeling Protocol

Klein, A., Tourville, J. 101 Labeled Brain Images and a Consistent Human Cortical Labeling Protocol. Frontiers in Neuroscience, vol 6 (2012)

work page 2012
[21]

An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest

Desikan, R.S., Ségonne, F., Fischl, B., Quinn, B.T., Dickerson, B.C., Blacker, D., Buckner, R.L., Dale, A.M., Maguire, R.P., Hyman, B.T., Albert, M.S., Killiany, R.J. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. NeuroImage, vol 31, issue 3 (2006)

work page 2006
[22]

N., Puonti, O., Thielscher, A., Van Leemput, K., Fischl, B., ..., Iglesias, J

Billot, B., Greve, D. N., Puonti, O., Thielscher, A., Van Leemput, K., Fischl, B., ..., Iglesias, J. E. SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining. Medical image analysis, 86, 102789 (2023)

work page 2023
[23]

FastSurfer - A fast and accurate deep learning based neuroimaging pipeline

Henschel, L., Conjeti, S., Estrada, S., Diers, L., Fischl, B., Reuter, M. FastSurfer - A fast and accurate deep learning based neuroimaging pipeline. NeuroImage, vol 219 (2020)

work page 2020
[24]

S., Evans, A

Fonov, V. S., Evans, A. C., McKinstry, R. C., Almli, C. R., Collins, D. L. Unbi- ased nonlinear average age-appropriate brain templates from birth to adulthood. NeuroImage, 47, S102 (2009)

work page 2009
[25]

F., Miller, M

Beg, M. F., Miller, M. I., Trouvé, A., Younes, L. Computing large deformation metric mappings via geodesic flows of diffeomorphisms. International journal of computer vision, 61, 139-157 (2005)

work page 2005

[1] [1]

D., Berendsen, F

De Vos, B. D., Berendsen, F. F., Viergever, M. A., Staring, M., Išgum, I. End- to-End Unsupervised Deformable Image Registration with a Convolutional Neural Network. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support LNCS, vol 10553. (2017)

work page 2017

[2] [2]

D., Berendsen, F

De Vos, B. D., Berendsen, F. F., Viergever, M. A., Sokooti, H., Staring, M., Išgum, I. A deep learning framework for unsupervised affine and deformable image regis- tration. Medical image analysis, 52, 128-143 (2019)

work page 2019

[3] [3]

Weakly-supervised convolutional neural networks for multimodal image registra- tion

Hu, Y., Modat, M., Gibson, E., Li, W., Ghavami, N., Bonmati, E., ..., Vercauteren, T. Weakly-supervised convolutional neural networks for multimodal image registra- tion. Medical image analysis, 49, 1-13 (2018)

work page 2018

[4] [4]

N., Iglesias, J

Hoffmann, M., Billot, B., Greve, D. N., Iglesias, J. E., Fischl, B., Dalca, A. V. Syn- thMorph: learning contrast-invariant registration without acquired images. Trans- actions on Medical Imaging, 41(3), 543-558 (2021)

work page 2021

[5] [5]

A ready-to-use machine learning tool for symmetric multi-modality registration of brain MRI

Iglesias, J.E. A ready-to-use machine learning tool for symmetric multi-modality registration of brain MRI. Scientific Reports, 13(1), 6657 (2023)

work page 2023

[6] [6]

M., Hu, Y

Baum, Z. M., Hu, Y. and Barratt, D. C. Multimodality biomedical image registra- tion using free point transformer networks. In Medical Ultrasound, and Preterm, Perinatal and Paediatric Image Analysis. First International Workshop, ASMUS 2020, PIPPI 2020, 116-125 (2020)

work page 2020

[7] [7]

M., Saeed, S

Min, Z., Baum, Z. M., Saeed, S. U., Emberton, M., Barratt, D. C., Taylor, Z. A. and Hu, Y. Biomechanics-informed Non-rigid Medical Image Registration and its Inverse Material Property Estimation with Linear and Nonlinear Elasticity. MICCAI 2024, 564-574 (2024)

work page 2024

[8] [8]

A Log-Euclidean Framework for Statistics on Diffeomorphisms

Arsigny, V., Commowick, O., Pennec, X., Ayache, N. A Log-Euclidean Framework for Statistics on Diffeomorphisms. MICCAI 2006, LNCS vol 4190 (2006)

work page 2006

[9] [9]

Contributions to 3D diffeomorphic atlas esti- mation: application to brain images

Bossa, M., Hernandez, M., Olmos, S. Contributions to 3D diffeomorphic atlas esti- mation: application to brain images. MICCAI, Proceedings, Part I 10, pp. 667-674 (2007)

work page 2007

[10] [10]

Symmetric log-domain dif- feomorphic registration: A demons-based approach

Vercauteren, T., Pennec, X., Perchant, A., Ayache, N. Symmetric log-domain dif- feomorphic registration: A demons-based approach. MICCAI, pp. 754-761 (2008)

work page 2008

[11] [11]

Unsupervised learning of probabilistic diffeomorphic registration for images and surfaces, Medical Image Analysis, vol 57, 226-236 (2019)

Dalca, A.V., Balakrishnan, G., Guttag, J., Sabuncu, M.R. Unsupervised learning of probabilistic diffeomorphic registration for images and surfaces, Medical Image Analysis, vol 57, 226-236 (2019)

work page 2019

[12] [12]

Yang, X., Li, Y., Reutens, D. et al. Diffeomorphic Metric Landmark Mapping Using Stationary Velocity Field Parameterization. Int J Comput Vis 115, 69–86 (2015)

work page 2015

[13] [13]

Lorensen, W. E. and Cline H. E. Marching cubes: A high resolution 3D surface construction algorithm. SIGGRAPH Comput. Graph. 21, 4, 163–169 (1987)

work page 1987

[14] [14]

POLAFFINI: Efficient Feature-Based Polyaffine Initialization for Improved Non-linear Image Registration

Legouhy, A., Callaghan, R., Azadbakht, H., Zhang, H. POLAFFINI: Efficient Feature-Based Polyaffine Initialization for Improved Non-linear Image Registration. IPMI 2023. LNCS, vol 13939 (2023)

work page 2023

[15] [15]

Arsigny, V., Commowick, O., Ayache, N. et al. A Fast and Log-Euclidean Polyaffine Framework for Locally Linear Registration. J Math Imaging Vis 33, 222–238 (2009). 10 A. Legouhy et al

work page 2009

[16] [16]

R., Su, H., Mo, K., Guibas, L.J

Qi, C. R., Su, H., Mo, K., Guibas, L.J. PointNet: deep learning on point sets for 3D classification and segmentation. CVPR, 652–660 (2017)

work page 2017

[17] [17]

A robust and efficient block matching framework for non linear registration of thoracic CT images

Garcia, V., Commowick, O., Malandain, G. A robust and efficient block matching framework for non linear registration of thoracic CT images. In Grand Challenges in Medical Image Analysis (MICCAI workshop) pp. 1-10 (2010)

work page 2010

[18] [18]

Besl, P. J. and McKay, N. D. Method for registration of 3-D shapes. Sensor fusion IV: control paradigms and data structures, vol. 1611, pp. 586-606 (1992)

work page 1992

[19] [19]

Alzheimer’s Disease Neuroimaging Initiative (ADNI): clinical char- acterization

Petersen, R.C., Aisen, P.S., Beckett, L.A., Donohue, M.C., Gamst, A.C., Harvey, D.J., Jack, C.R., Jr, Jagust, W.J., Shaw, L.M., Toga, A.W., Trojanowski, J.Q., Weiner, M.W. Alzheimer’s Disease Neuroimaging Initiative (ADNI): clinical char- acterization. Neurology, 74(3), 201–209 (2010)

work page 2010

[20] [20]

101 Labeled Brain Images and a Consistent Human Cortical Labeling Protocol

Klein, A., Tourville, J. 101 Labeled Brain Images and a Consistent Human Cortical Labeling Protocol. Frontiers in Neuroscience, vol 6 (2012)

work page 2012

[21] [21]

An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest

Desikan, R.S., Ségonne, F., Fischl, B., Quinn, B.T., Dickerson, B.C., Blacker, D., Buckner, R.L., Dale, A.M., Maguire, R.P., Hyman, B.T., Albert, M.S., Killiany, R.J. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. NeuroImage, vol 31, issue 3 (2006)

work page 2006

[22] [22]

N., Puonti, O., Thielscher, A., Van Leemput, K., Fischl, B., ..., Iglesias, J

Billot, B., Greve, D. N., Puonti, O., Thielscher, A., Van Leemput, K., Fischl, B., ..., Iglesias, J. E. SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining. Medical image analysis, 86, 102789 (2023)

work page 2023

[23] [23]

FastSurfer - A fast and accurate deep learning based neuroimaging pipeline

Henschel, L., Conjeti, S., Estrada, S., Diers, L., Fischl, B., Reuter, M. FastSurfer - A fast and accurate deep learning based neuroimaging pipeline. NeuroImage, vol 219 (2020)

work page 2020

[24] [24]

S., Evans, A

Fonov, V. S., Evans, A. C., McKinstry, R. C., Almli, C. R., Collins, D. L. Unbi- ased nonlinear average age-appropriate brain templates from birth to adulthood. NeuroImage, 47, S102 (2009)

work page 2009

[25] [25]

F., Miller, M

Beg, M. F., Miller, M. I., Trouvé, A., Younes, L. Computing large deformation metric mappings via geodesic flows of diffeomorphisms. International journal of computer vision, 61, 139-157 (2005)

work page 2005