-
Label merge-and-split: A graph-colouring approach for memory-efficient brain parcellation
Authors:
Aaron Kujawa,
Reuben Dorent,
Sebastien Ourselin,
Tom Vercauteren
Abstract:
Whole brain parcellation requires inferring hundreds of segmentation labels in large image volumes and thus presents significant practical challenges for deep learning approaches. We introduce label merge-and-split, a method that first greatly reduces the effective number of labels required for learning-based whole brain parcellation and then recovers original labels. Using a greedy graph colourin…
▽ More
Whole brain parcellation requires inferring hundreds of segmentation labels in large image volumes and thus presents significant practical challenges for deep learning approaches. We introduce label merge-and-split, a method that first greatly reduces the effective number of labels required for learning-based whole brain parcellation and then recovers original labels. Using a greedy graph colouring algorithm, our method automatically groups and merges multiple spatially separate labels prior to model training and inference. The merged labels may be semantically unrelated. A deep learning model is trained to predict merged labels. At inference time, original labels are restored using atlas-based influence regions. In our experiments, the proposed approach reduces the number of labels by up to 68% while achieving segmentation accuracy comparable to the baseline method without label merging and splitting. Moreover, model training and inference times as well as GPU memory requirements were reduced significantly. The proposed method can be applied to all semantic segmentation tasks with a large number of spatially separate classes within an atlas-based prior.
△ Less
Submitted 1 August, 2024; v1 submitted 16 April, 2024;
originally announced April 2024.
-
MedShapeNet -- A Large-Scale Dataset of 3D Medical Shapes for Computer Vision
Authors:
Jianning Li,
Zongwei Zhou,
Jiancheng Yang,
Antonio Pepe,
Christina Gsaxner,
Gijs Luijten,
Chongyu Qu,
Tiezheng Zhang,
Xiaoxi Chen,
Wenxuan Li,
Marek Wodzinski,
Paul Friedrich,
Kangxian Xie,
Yuan Jin,
Narmada Ambigapathy,
Enrico Nasca,
Naida Solak,
Gian Marco Melito,
Viet Duc Vu,
Afaque R. Memon,
Christopher Schlachta,
Sandrine De Ribaupierre,
Rajnikant Patel,
Roy Eagleson,
Xiaojun Chen
, et al. (132 additional authors not shown)
Abstract:
Prior to the deep learning era, shape was commonly used to describe the objects. Nowadays, state-of-the-art (SOTA) algorithms in medical imaging are predominantly diverging from computer vision, where voxel grids, meshes, point clouds, and implicit surface models are used. This is seen from numerous shape-related publications in premier vision conferences as well as the growing popularity of Shape…
▽ More
Prior to the deep learning era, shape was commonly used to describe the objects. Nowadays, state-of-the-art (SOTA) algorithms in medical imaging are predominantly diverging from computer vision, where voxel grids, meshes, point clouds, and implicit surface models are used. This is seen from numerous shape-related publications in premier vision conferences as well as the growing popularity of ShapeNet (about 51,300 models) and Princeton ModelNet (127,915 models). For the medical domain, we present a large collection of anatomical shapes (e.g., bones, organs, vessels) and 3D models of surgical instrument, called MedShapeNet, created to facilitate the translation of data-driven vision algorithms to medical applications and to adapt SOTA vision algorithms to medical problems. As a unique feature, we directly model the majority of shapes on the imaging data of real patients. As of today, MedShapeNet includes 23 dataset with more than 100,000 shapes that are paired with annotations (ground truth). Our data is freely accessible via a web interface and a Python application programming interface (API) and can be used for discriminative, reconstructive, and variational benchmarks as well as various applications in virtual, augmented, or mixed reality, and 3D printing. Exemplary, we present use cases in the fields of classification of brain tumors, facial and skull reconstructions, multi-class anatomy completion, education, and 3D printing. In future, we will extend the data and improve the interfaces. The project pages are: https://medshapenet.ikim.nrw/ and https://github.com/Jianningli/medshapenet-feedback
△ Less
Submitted 12 December, 2023; v1 submitted 30 August, 2023;
originally announced August 2023.
-
Boundary Distance Loss for Intra-/Extra-meatal Segmentation of Vestibular Schwannoma
Authors:
Navodini Wijethilake,
Aaron Kujawa,
Reuben Dorent,
Muhammad Asad,
Anna Oviedova,
Tom Vercauteren,
Jonathan Shapey
Abstract:
Vestibular Schwannoma (VS) typically grows from the inner ear to the brain. It can be separated into two regions, intrameatal and extrameatal respectively corresponding to being inside or outside the inner ear canal. The growth of the extrameatal regions is a key factor that determines the disease management followed by the clinicians. In this work, a VS segmentation approach with subdivision into…
▽ More
Vestibular Schwannoma (VS) typically grows from the inner ear to the brain. It can be separated into two regions, intrameatal and extrameatal respectively corresponding to being inside or outside the inner ear canal. The growth of the extrameatal regions is a key factor that determines the disease management followed by the clinicians. In this work, a VS segmentation approach with subdivision into intra-/extra-meatal parts is presented. We annotated a dataset consisting of 227 T2 MRI instances, acquired longitudinally on 137 patients, excluding post-operative instances. We propose a staged approach, with the first stage performing the whole tumour segmentation and the second stage performing the intra-/extra-meatal segmentation using the T2 MRI along with the mask obtained from the first stage. To improve on the accuracy of the predicted meatal boundary, we introduce a task-specific loss which we call Boundary Distance Loss. The performance is evaluated in contrast to the direct intrameatal extrameatal segmentation task performance, i.e. the Baseline. Our proposed method, with the two-stage approach and the Boundary Distance Loss, achieved a Dice score of 0.8279+-0.2050 and 0.7744+-0.1352 for extrameatal and intrameatal regions respectively, significantly improving over the Baseline, which gave Dice score of 0.7939+-0.2325 and 0.7475+-0.1346 for the extrameatal and intrameatal regions respectively.
△ Less
Submitted 9 August, 2022;
originally announced August 2022.
-
CrossMoDA 2021 challenge: Benchmark of Cross-Modality Domain Adaptation techniques for Vestibular Schwannoma and Cochlea Segmentation
Authors:
Reuben Dorent,
Aaron Kujawa,
Marina Ivory,
Spyridon Bakas,
Nicola Rieke,
Samuel Joutard,
Ben Glocker,
Jorge Cardoso,
Marc Modat,
Kayhan Batmanghelich,
Arseniy Belkov,
Maria Baldeon Calisto,
Jae Won Choi,
Benoit M. Dawant,
Hexin Dong,
Sergio Escalera,
Yubo Fan,
Lasse Hansen,
Mattias P. Heinrich,
Smriti Joshi,
Victoriya Kashtanova,
Hyeon Gyu Kim,
Satoshi Kondo,
Christian N. Kruse,
Susana K. Lai-Yuen
, et al. (15 additional authors not shown)
Abstract:
Domain Adaptation (DA) has recently raised strong interests in the medical imaging community. While a large variety of DA techniques has been proposed for image segmentation, most of these techniques have been validated either on private datasets or on small publicly available datasets. Moreover, these datasets mostly addressed single-class problems. To tackle these limitations, the Cross-Modality…
▽ More
Domain Adaptation (DA) has recently raised strong interests in the medical imaging community. While a large variety of DA techniques has been proposed for image segmentation, most of these techniques have been validated either on private datasets or on small publicly available datasets. Moreover, these datasets mostly addressed single-class problems. To tackle these limitations, the Cross-Modality Domain Adaptation (crossMoDA) challenge was organised in conjunction with the 24th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2021). CrossMoDA is the first large and multi-class benchmark for unsupervised cross-modality DA. The challenge's goal is to segment two key brain structures involved in the follow-up and treatment planning of vestibular schwannoma (VS): the VS and the cochleas. Currently, the diagnosis and surveillance in patients with VS are performed using contrast-enhanced T1 (ceT1) MRI. However, there is growing interest in using non-contrast sequences such as high-resolution T2 (hrT2) MRI. Therefore, we created an unsupervised cross-modality segmentation benchmark. The training set provides annotated ceT1 (N=105) and unpaired non-annotated hrT2 (N=105). The aim was to automatically perform unilateral VS and bilateral cochlea segmentation on hrT2 as provided in the testing set (N=137). A total of 16 teams submitted their algorithm for the evaluation phase. The level of performance reached by the top-performing teams is strikingly high (best median Dice - VS:88.4%; Cochleas:85.7%) and close to full supervision (median Dice - VS:92.5%; Cochleas:87.7%). All top-performing methods made use of an image-to-image translation approach to transform the source-domain images into pseudo-target-domain images. A segmentation network was then trained using these generated images and the manual annotations provided for the source image.
△ Less
Submitted 14 December, 2022; v1 submitted 8 January, 2022;
originally announced January 2022.
-
Statistical Review of Animal trials -- A Guideline
Authors:
Sophie K. Piper,
Dario Zocholl,
Ulf Toelch,
Robert Roehle,
Andrea Stroux,
Johanna Hößler,
Anne Kujawa,
Frank Konietschke
Abstract:
Examinations of any experiment involving living organisms require justifications of the need and moral defensibleness of the study. Statistical planning, design and sample size calculation of the experiment are no less important review criteria than general medical and ethical points to consider. Errors made in the statistical planning and data evaluation phase can have severe consequences on both…
▽ More
Examinations of any experiment involving living organisms require justifications of the need and moral defensibleness of the study. Statistical planning, design and sample size calculation of the experiment are no less important review criteria than general medical and ethical points to consider. Errors made in the statistical planning and data evaluation phase can have severe consequences on both results and conclusions. They might proliferate and thus impact future trials-an unintended outcome of fundamental research with profound ethical consequences. Therefore, any trial must be efficient in both a medical and statistical way in answering the questions of interests to be considered as approvable. Unified statistical standards are currently missing for animal review boards in Germany. In order to accompany, we developed a biometric form to be filled and handed in with the proposal at the local authority on animal welfare. It addresses relevant points to consider for biostatistical planning of animal experiments and can help both the applicants and the reviewers in overseeing the entire experiment(s) planned. Furthermore, the form might also aid in meeting the current standards set by the 3+3R's principle of animal experimentation Replacement, Reduction, Refinement as well as Robustness, Registration and Reporting. The form has already been in use by the local authority of animal welfare in Berlin, Germany. In addition, we provide reference to our user guide giving more detailed explanation and examples for each section of the biometric form. Unifying the set of biostatistical aspects will help both the applicants and the reviewers to equal standards and increase quality of preclinical research projects, also for translational, multicenter, or international studies.
△ Less
Submitted 21 November, 2021; v1 submitted 29 October, 2021;
originally announced October 2021.
-
Inter Extreme Points Geodesics for End-to-End Weakly Supervised Image Segmentation
Authors:
Reuben Dorent,
Samuel Joutard,
Jonathan Shapey,
Aaron Kujawa,
Marc Modat,
Sebastien Ourselin,
Tom Vercauteren
Abstract:
We introduce $\textit{InExtremIS}$, a weakly supervised 3D approach to train a deep image segmentation network using particularly weak train-time annotations: only 6 extreme clicks at the boundary of the objects of interest. Our fully-automatic method is trained end-to-end and does not require any test-time annotations. From the extreme points, 3D bounding boxes are extracted around objects of int…
▽ More
We introduce $\textit{InExtremIS}$, a weakly supervised 3D approach to train a deep image segmentation network using particularly weak train-time annotations: only 6 extreme clicks at the boundary of the objects of interest. Our fully-automatic method is trained end-to-end and does not require any test-time annotations. From the extreme points, 3D bounding boxes are extracted around objects of interest. Then, deep geodesics connecting extreme points are generated to increase the amount of "annotated" voxels within the bounding boxes. Finally, a weakly supervised regularised loss derived from a Conditional Random Field formulation is used to encourage prediction consistency over homogeneous regions. Extensive experiments are performed on a large open dataset for Vestibular Schwannoma segmentation. $\textit{InExtremIS}$ obtained competitive performance, approaching full supervision and outperforming significantly other weakly supervised techniques based on bounding boxes. Moreover, given a fixed annotation time budget, $\textit{InExtremIS}$ outperforms full supervision. Our code and data are available online.
△ Less
Submitted 4 November, 2021; v1 submitted 1 July, 2021;
originally announced July 2021.
-
A comparison of the cut-off effects for Twisted Mass, Overlap and Creutz fermions at tree-level of Perturbation Theory
Authors:
Krzysztof Cichy,
Jenifer Gonzalez Lopez,
Agnieszka Kujawa
Abstract:
In this paper we investigate the cutoff effects at tree-level of perturbation theory for three different lattice regularizations of fermions -- maximally twisted mass Wilson, overlap and Creutz fermions. We show that all three kinds of fermions exhibit the expected $O(a^2)$ scaling behaviour in the lattice spacing. Moreover, the size of these cutoff effects for the considered quantities i.e. the…
▽ More
In this paper we investigate the cutoff effects at tree-level of perturbation theory for three different lattice regularizations of fermions -- maximally twisted mass Wilson, overlap and Creutz fermions. We show that all three kinds of fermions exhibit the expected $O(a^2)$ scaling behaviour in the lattice spacing. Moreover, the size of these cutoff effects for the considered quantities i.e. the pseudoscalar correlation function $C_{PS}$, the mass $m_{PS}$ and the decay constant $f_{PS}$ is comparable for all of them.
△ Less
Submitted 4 November, 2008;
originally announced November 2008.
-
Some Properties of the Spin-Polarized Attractive Hubbard Model
Authors:
Agnieszka Kujawa,
Roman Micnas
Abstract:
The influence of a Zeeman magnetic field on the superconducting characteristics of the attractive Hubbard model was investigated. The ground state and temperature phase diagrams were obtained for a fixed number of particles. Two critical magnetic fields were found for the first order phase transition from the superconducting to the normal state for $n\neq 1$. For some range of parameters a reent…
▽ More
The influence of a Zeeman magnetic field on the superconducting characteristics of the attractive Hubbard model was investigated. The ground state and temperature phase diagrams were obtained for a fixed number of particles. Two critical magnetic fields were found for the first order phase transition from the superconducting to the normal state for $n\neq 1$. For some range of parameters a reentrant transition was found and gapless superconductivity was obtained. The significance of the Hartree term was also analyzed.
△ Less
Submitted 24 February, 2009; v1 submitted 8 September, 2008;
originally announced September 2008.
-
On the Phase Diagram of Spin-Polarized Attractive Hubbard Model: Weak Coupling Limit
Authors:
Agnieszka Kujawa,
Roman Micnas
Abstract:
The superfluid properties of the attractive Hubbard model in a Zeeman magnetic field, and in the weak coupling regime have been investigated. The temperature and magnetic field dependencies of the order parameter have been analyzed. Furthermore, the temperature vs. magnetic field and temperature vs. spin polarization phase diagrams for the 2D and 3D lattices have been obtained. For some paramete…
▽ More
The superfluid properties of the attractive Hubbard model in a Zeeman magnetic field, and in the weak coupling regime have been investigated. The temperature and magnetic field dependencies of the order parameter have been analyzed. Furthermore, the temperature vs. magnetic field and temperature vs. spin polarization phase diagrams for the 2D and 3D lattices have been obtained. For some parameters a reentrant transition has been found.
△ Less
Submitted 11 August, 2008;
originally announced August 2008.
-
The BCS-BEC crossover and superconductivity in a lattice fermion model with hard core repulsion
Authors:
Agnieszka Kujawa
Abstract:
We have considered the BCS-BEC crossover at T=0 in the framework of a lattice fermion model (the extended Hubbard model). We have analyzed the case of on-site hard core repulsion ($U=\infty$) and intersite attraction ($W<0$) in a 3D system with extended s-wave pairing symmetry. The mechanisms of density-driven and attraction-strength-driven crossover have been analyzed. We have obtained the grou…
▽ More
We have considered the BCS-BEC crossover at T=0 in the framework of a lattice fermion model (the extended Hubbard model). We have analyzed the case of on-site hard core repulsion ($U=\infty$) and intersite attraction ($W<0$) in a 3D system with extended s-wave pairing symmetry. The mechanisms of density-driven and attraction-strength-driven crossover have been analyzed. We have obtained the ground state phase diagram. The analyzed quantities are: chemical potential, order parameter, coherence length and fraction of condensed particles.
△ Less
Submitted 11 August, 2008;
originally announced August 2008.
-
Twisted Mass, Overlap and Creutz Fermions: Cut-off Effects at Tree-level of Perturbation Theory
Authors:
K. Cichy,
J. Gonzalez Lopez,
K. Jansen,
A. Kujawa,
A. Shindler
Abstract:
We study cutoff effects at tree-level of perturbation theory for maximally twisted mass Wilson, overlap and the recently proposed Creutz fermions. We demonstrate that all three kind of lattice fermions exhibit the expected O(a^2) scaling behaviour in the lattice spacing. In addition, the sizes of these cutoff effects are comparable for the three kinds of lattice fermions considered here. Further…
▽ More
We study cutoff effects at tree-level of perturbation theory for maximally twisted mass Wilson, overlap and the recently proposed Creutz fermions. We demonstrate that all three kind of lattice fermions exhibit the expected O(a^2) scaling behaviour in the lattice spacing. In addition, the sizes of these cutoff effects are comparable for the three kinds of lattice fermions considered here. Furthermore, we analyze situations when twisted mass fermions are not exactly at maximal twist and when overlap fermions are studied in comparison to twisted mass fermions when the quark masses are not matched.
△ Less
Submitted 6 March, 2008; v1 submitted 25 February, 2008;
originally announced February 2008.
-
Cutoff effects for Wilson twisted mass fermions at tree-level of perturbation theory
Authors:
K. Cichy,
J. Gonzalez Lopez,
K. Jansen,
A. Kujawa,
A. Shindler
Abstract:
We study cutoff effects at tree-level of perturbation theory for standard Wilson and Wilson twisted mass fermionic lattice actions with Nf=2 flavour degenerate quarks. The discretization effects are investigated by computing the mass spectrum and decay amplitudes for different hadron interpolating fields and the scaling behaviour towards the continuum limit is analyzed. It is shown that the Wils…
▽ More
We study cutoff effects at tree-level of perturbation theory for standard Wilson and Wilson twisted mass fermionic lattice actions with Nf=2 flavour degenerate quarks. The discretization effects are investigated by computing the mass spectrum and decay amplitudes for different hadron interpolating fields and the scaling behaviour towards the continuum limit is analyzed. It is shown that the Wilson and the mass average methods are equivalent and lead to O(a) improved R5-parity even lattice observables. We also demonstrate that automatic O(a) improvement works in case of Wilson twisted mass fermions at maximal twist and that this improvement is realized even if the condition of maximal twist is achieved only up to O(a) cutoff effects. We demonstrate that in the chiral limit standard Wilson fermions show scaling violations of O(a^2) while for maximally twisted mass fermions these violations are only of O(a^4). For our analytical calculations, lattices with sizes L=aN and periodic boundary conditions in the spatial directions have been chosen while infinite extension in the time direction is considered.
△ Less
Submitted 10 October, 2007;
originally announced October 2007.