-
Anatomy-informed Data Augmentation for Enhanced Prostate Cancer Detection
Authors:
Balint Kovacs,
Nils Netzer,
Michael Baumgartner,
Carolin Eith,
Dimitrios Bounias,
Clara Meinzer,
Paul F. Jaeger,
Kevin S. Zhang,
Ralf Floca,
Adrian Schrader,
Fabian Isensee,
Regula Gnirs,
Magdalena Goertz,
Viktoria Schuetz,
Albrecht Stenzinger,
Markus Hohenfellner,
Heinz-Peter Schlemmer,
Ivo Wolf,
David Bonekamp,
Klaus H. Maier-Hein
Abstract:
Data augmentation (DA) is a key factor in medical image analysis, such as in prostate cancer (PCa) detection on magnetic resonance images. State-of-the-art computer-aided diagnosis systems still rely on simplistic spatial transformations to preserve the pathological label post transformation. However, such augmentations do not substantially increase the organ as well as tumor shape variability in…
▽ More
Data augmentation (DA) is a key factor in medical image analysis, such as in prostate cancer (PCa) detection on magnetic resonance images. State-of-the-art computer-aided diagnosis systems still rely on simplistic spatial transformations to preserve the pathological label post transformation. However, such augmentations do not substantially increase the organ as well as tumor shape variability in the training set, limiting the model's ability to generalize to unseen cases with more diverse localized soft-tissue deformations. We propose a new anatomy-informed transformation that leverages information from adjacent organs to simulate typical physiological deformations of the prostate and generates unique lesion shapes without altering their label. Due to its lightweight computational requirements, it can be easily integrated into common DA frameworks. We demonstrate the effectiveness of our augmentation on a dataset of 774 biopsy-confirmed examinations, by evaluating a state-of-the-art method for PCa detection with different augmentation settings.
△ Less
Submitted 7 September, 2023;
originally announced September 2023.
-
The Torus of Triangles
Authors:
Eric Brussel,
Madeleine E. Goertz
Abstract:
We show that when parameterized by triples of angles, the set of similarity classes of labeled, oriented, possibly degenerate triangles has the natural structure of the Clifford torus $\sf{T}$, a compact abelian Lie group. On this torus the main triangle types form distinguished algebraic structures: subgroups and cosets. The construction relies on a natural definition of similarity for degenerate…
▽ More
We show that when parameterized by triples of angles, the set of similarity classes of labeled, oriented, possibly degenerate triangles has the natural structure of the Clifford torus $\sf{T}$, a compact abelian Lie group. On this torus the main triangle types form distinguished algebraic structures: subgroups and cosets. The construction relies on a natural definition of similarity for degenerate triangles.
We analyze the set of (unrestricted) similarity classes using a uniform probability measure on $\sf{T}$ and compute the relative measures of the different triangle types.
Our computations are compatible with the spherically symmetric probability distribution analyzed in [Por94] and [ES15], which are based on vertices/side lengths instead of angles.
△ Less
Submitted 16 August, 2023; v1 submitted 20 March, 2023;
originally announced March 2023.
-
Multiscale methods for solving wave equations on spatial networks
Authors:
Morgan Görtz,
Per Ljung,
Axel Målqvist
Abstract:
We present and analyze a multiscale method for wave propagation problems, posed on spatial networks. By introducing a coarse scale, using a finite element space interpolated onto the network, we construct a discrete multiscale space using the localized orthogonal decomposition (LOD) methodology. The spatial discretization is then combined with an energy conserving temporal scheme to form the propo…
▽ More
We present and analyze a multiscale method for wave propagation problems, posed on spatial networks. By introducing a coarse scale, using a finite element space interpolated onto the network, we construct a discrete multiscale space using the localized orthogonal decomposition (LOD) methodology. The spatial discretization is then combined with an energy conserving temporal scheme to form the proposed method. Under the assumption of well-prepared initial data, we derive an a priori error bound of optimal order with respect to the space and time discretization. In the analysis, we combine the theory derived for stationary elliptic problems on spatial networks with classical finite element results for hyperbolic problems. Finally, we present numerical experiments that confirm our theoretical findings.
△ Less
Submitted 15 March, 2023; v1 submitted 25 November, 2022;
originally announced November 2022.
-
Numerical homogenization of spatial network models
Authors:
Fredrik Edelvik,
Morgan Görtz,
Fredrik Hellman,
Gustav Kettil,
Axel Målqvist
Abstract:
We present and analyze a methodology for numerical homogenization of spatial networks, modelling e.g. diffusion processes and deformation of mechanical structures. The aim is to construct an accurate coarse model of the network. By solving decoupled problems on local subgraphs we construct a low dimensional subspace of the solution space with good approximation properties. The coarse model of the…
▽ More
We present and analyze a methodology for numerical homogenization of spatial networks, modelling e.g. diffusion processes and deformation of mechanical structures. The aim is to construct an accurate coarse model of the network. By solving decoupled problems on local subgraphs we construct a low dimensional subspace of the solution space with good approximation properties. The coarse model of the network is expressed by a Galerkin formulation and can be used to perform simulations with different source and boundary data at a low computational cost. We prove optimal convergence of the proposed method under mild assumptions on the homogeneity, connectivity, and locality of the network on the coarse scale. The theoretical findings are numerically confirmed for both scalar-valued (heat conduction) and vector-valued (structural) models.
△ Less
Submitted 13 September, 2022;
originally announced September 2022.
-
Iterative solution of spatial network models by subspace decomposition
Authors:
Morgan Görtz,
Fredrik Hellman,
Axel Målqvist
Abstract:
We present and analyze a preconditioned conjugate gradient method (PCG) for solving spatial network problems. Primarily, we consider diffusion and structural mechanics simulations for fiber based materials, but the methodology can be applied to a wide range of models, fulfilling a set of abstract assumptions. The proposed method builds on a classical subspace decomposition into a coarse subspace,…
▽ More
We present and analyze a preconditioned conjugate gradient method (PCG) for solving spatial network problems. Primarily, we consider diffusion and structural mechanics simulations for fiber based materials, but the methodology can be applied to a wide range of models, fulfilling a set of abstract assumptions. The proposed method builds on a classical subspace decomposition into a coarse subspace, realized as the restriction of a finite element space to the nodes of the spatial network, and localized subspaces with support on mesh stars. The main contribution of this work is the convergence analysis of the proposed method. The analysis translates results from finite element theory, including interpolation bounds, to the spatial network setting. A convergence rate of the PCG algorithm, only depending on global bounds of the operator and homogeneity, connectivity and locality constants of the network, is established. The theoretical results are confirmed by several numerical experiments.
△ Less
Submitted 15 July, 2022;
originally announced July 2022.
-
Nonlinear dynamics of an epidemic compartment model with asymptomatic infections and mitigation
Authors:
Maurice Görtz,
Joachim Krug
Abstract:
A significant proportion of the infections driving the current {SARS-CoV-2} pandemic are transmitted asymptomatically. Here we introduce and study a simple epidemic model with separate compartments comprising asymptomatic and symptomatic infected individuals. The linear dynamics determining the outbreak condition of the model is equivalent to a renewal theory approach with exponential waiting time…
▽ More
A significant proportion of the infections driving the current {SARS-CoV-2} pandemic are transmitted asymptomatically. Here we introduce and study a simple epidemic model with separate compartments comprising asymptomatic and symptomatic infected individuals. The linear dynamics determining the outbreak condition of the model is equivalent to a renewal theory approach with exponential waiting time distributions. Exploiting a nontrivial conservation law of the full nonlinear dynamics, we derive analytic bounds on the peak number of infections in the absence and presence of mitigation through isolation and testing. The bounds are compared to numerical solutions of the differential equations.
△ Less
Submitted 8 October, 2022; v1 submitted 12 April, 2022;
originally announced April 2022.
-
A numerical multiscale method for fiber networks
Authors:
Morgan Görtz,
Gustav Kettil,
Axel Målqvist,
Andreas Mark,
Fredrik Edelvik
Abstract:
Fiber network modeling can be used for studying mechanical properties of paper. The individual fibers and the bonds in-between constitute a detailed representation of the material. However, detailed microscale fiber network models must be resolved with efficient numerical methods. In this work, a numerical multiscale method for discrete network models is proposed that is based on the localized ort…
▽ More
Fiber network modeling can be used for studying mechanical properties of paper. The individual fibers and the bonds in-between constitute a detailed representation of the material. However, detailed microscale fiber network models must be resolved with efficient numerical methods. In this work, a numerical multiscale method for discrete network models is proposed that is based on the localized orthogonal decomposition method [arXiv:1810.05059]. The method is ideal for these network problems, because it reduces the maximum size of the problem, it is suitable for parallelization, and it can effectively solve fracture propagation.
The problem analyzed in this work is the nodal displacement of a fiber network given an applied load. This problem is formulated as a linear system that is solved by using the aforementioned multiscale method. To solve the linear system, the multiscale method constructs a low-dimensional solution space with good approximation properties. The method is observed to work well for unstructured fiber networks, with optimal rates of convergence obtainable for highly localized configurations of the method.
△ Less
Submitted 28 April, 2020;
originally announced April 2020.
