Showing 1–8 of 8 results for author: Schott, B

Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: Evaluation in Alzheimer's disease

Authors: Martin Dyrba, Moritz Hanzig, Slawek Altenstein, Sebastian Bader, Tommaso Ballarini, Frederic Brosseron, Katharina Buerger, Daniel Cantré, Peter Dechent, Laura Dobisch, Emrah Düzel, Michael Ewers, Klaus Fliessbach, Wenzel Glanz, John-Dylan Haynes, Michael T. Heneka, Daniel Janowitz, Deniz B. Keles, Ingo Kilimann, Christoph Laske, Franziska Maier, Coraline D. Metzger, Matthias H. Munk, Robert Perneczky, Oliver Peters , et al. (15 additional authors not shown)

Abstract: Background: Although convolutional neural networks (CNN) achieve high diagnostic accuracy for detecting Alzheimer's disease (AD) dementia based on magnetic resonance imaging (MRI) scans, they are not yet applied in clinical routine. One important reason for this is a lack of model comprehensibility. Recently developed visualization methods for deriving CNN relevance maps may help to fill this gap.… ▽ More Background: Although convolutional neural networks (CNN) achieve high diagnostic accuracy for detecting Alzheimer's disease (AD) dementia based on magnetic resonance imaging (MRI) scans, they are not yet applied in clinical routine. One important reason for this is a lack of model comprehensibility. Recently developed visualization methods for deriving CNN relevance maps may help to fill this gap. We investigated whether models with higher accuracy also rely more on discriminative brain regions predefined by prior knowledge. Methods: We trained a CNN for the detection of AD in N=663 T1-weighted MRI scans of patients with dementia and amnestic mild cognitive impairment (MCI) and verified the accuracy of the models via cross-validation and in three independent samples including N=1655 cases. We evaluated the association of relevance scores and hippocampus volume to validate the clinical utility of this approach. To improve model comprehensibility, we implemented an interactive visualization of 3D CNN relevance maps. Results: Across three independent datasets, group separation showed high accuracy for AD dementia vs. controls (AUC$\geq$0.92) and moderate accuracy for MCI vs. controls (AUC$\approx$0.75). Relevance maps indicated that hippocampal atrophy was considered as the most informative factor for AD detection, with additional contributions from atrophy in other cortical and subcortical regions. Relevance scores within the hippocampus were highly correlated with hippocampal volumes (Pearson's r$\approx$-0.86, p<0.001). Conclusion: The relevance maps highlighted atrophy in regions that we had hypothesized a priori. This strengthens the comprehensibility of the CNN models, which were trained in a purely data-driven manner based on the scans and diagnosis labels. △ Less

Submitted 5 November, 2021; v1 submitted 18 December, 2020; originally announced December 2020.

Comments: 24 pages, 9 figures/tables, supplementary material, source code available on GitHub

A consistent approach for fluid-structure-contact interaction based on a porous flow model for rough surface contact

Authors: Christoph Ager, Benedikt Schott, Anh-Tu Vuong, Alexander Popp, Wolfgang A. Wall

Abstract: Simulation approaches for fluid-structure-contact interaction, especially if requested to be consistent even down to the real contact scenarios, belong to the most challenging and still unsolved problems in computational mechanics. The main challenges are twofold - one is to have a correct physical model for this scenario, and the other one is to have a numerical method that is capable of working… ▽ More Simulation approaches for fluid-structure-contact interaction, especially if requested to be consistent even down to the real contact scenarios, belong to the most challenging and still unsolved problems in computational mechanics. The main challenges are twofold - one is to have a correct physical model for this scenario, and the other one is to have a numerical method that is capable of working and being consistent down to a zero gap. And when analyzing such challenging setups of fluid-structure interaction that include contact of submersed solid components it gets obvious that the influence of surface roughness effects is essential for a physical consistent modeling of such configurations. To capture this system behavior, we present a continuum mechanical model which is able to include the effects of the surface microstructure in a fluid-structure-contact interaction framework. An averaged representation for the mixture of fluid and solid on the rough surfaces, which is of major interest for the macroscopic response of such a system, is introduced therein. The inherent coupling of the macroscopic fluid flow and the flow inside the rough surfaces, the stress exchange of all contacting solid bodies involved, and the interaction between fluid and solid is included in the construction of the model. Although the physical model is not restricted to finite element based methods, a numerical approach with its core based on the Cut Finite Element Method (CutFEM), enabling topological changes of the fluid domain to solve the presented model numerically, is introduced. Such a CutFEM based approach is able to deal with the numerical challenges mentioned above. Different test cases give a perspective towards the potential capabilities of the presented physical model and numerical approach. △ Less

Submitted 11 September, 2018; originally announced September 2018.

Comments: 33 pages, 23 figures

A Nitsche-based cut finite element method for the coupling of incompressible fluid flow with poroelasticity

Authors: Christoph Ager, Benedikt Schott, Magnus Winter, Wolfgang A. Wall

Abstract: The focus of this contribution is the numerical treatment of interface coupled problems concerning the interaction of incompressible fluid flow and permeable, elastic structures. The main emphasis is on extending the range of applicability of available formulations on especially three aspects. These aspects are the incorporation of a more general poroelasticity formulation, the use of the cut fini… ▽ More The focus of this contribution is the numerical treatment of interface coupled problems concerning the interaction of incompressible fluid flow and permeable, elastic structures. The main emphasis is on extending the range of applicability of available formulations on especially three aspects. These aspects are the incorporation of a more general poroelasticity formulation, the use of the cut finite element method (CutFEM) to allow for large interface motion and topological changes of the fluid domain, and the application of a novel Nitsche-based approach to incorporate the Beavers-Joseph (-Saffmann) tangential interface condition. This last aspect allows one to extend the practicable range of applicability of the proposed formulation down to very low porosities and permeabilities which is important in several examples in application. Different aspects of the presented formulation are analyzed in a numerical example including spatial convergence, the sensitivity of the solution to the Nitsche penalty parameters, varying porosities and permeabilities, and a varying Beavers-Joseph interface model constant. Finally, a numerical example analyzing the fluid induced bending of a poroelastic beam provides evidence of the general applicability of the presented approach. △ Less

Submitted 17 August, 2018; originally announced August 2018.

Comments: 25 pages, 13 figures

A monolithic approach to fluid-structure interaction based on a hybrid Eulerian-ALE fluid domain decomposition involving cut elements

Authors: Benedikt Schott, Christoph Ager, Wolfgang A. Wall

Abstract: A novel method for complex fluid-structure interaction (FSI) involving large structural deformation and motion is proposed. The new approach is based on a hybrid fluid formulation that combines the advantages of purely Eulerian (fixed-grid) and Arbitrary-Lagrangian-Eulerian (ALE moving mesh) formulations in the context of FSI. The structure - as commonly given in Lagrangian description - is surrou… ▽ More A novel method for complex fluid-structure interaction (FSI) involving large structural deformation and motion is proposed. The new approach is based on a hybrid fluid formulation that combines the advantages of purely Eulerian (fixed-grid) and Arbitrary-Lagrangian-Eulerian (ALE moving mesh) formulations in the context of FSI. The structure - as commonly given in Lagrangian description - is surrounded by a fine resolved layer of fluid elements based on an ALE-framework. This ALE-fluid patch, which is embedded in an Eulerian background fluid domain, follows the deformation and motion of the structural interface. This approximation technique is not limited to Finite Element Methods, but can can also be realized within other frameworks like Finite Volume or Discontinuous Galerkin Methods. In this work, the surface coupling between the two disjoint fluid subdomains is imposed weakly using a stabilized Nitsche's technique in a Cut Finite Element Method (CutFEM) framework. At the fluid-solid interface, standard weak coupling of node-matching or non-matching finite element approximations can be utilized. As the fluid subdomains can be meshed independently, a sufficient mesh quality in the vicinity of the common fluid-structure interface can be assured. To our knowledge the proposed method is the only method (despite some overlap** domain decomposition approaches that suffer from other issues) that allows for capturing boundary layers and flow detachment via appropriate grids around largely moving and deforming bodies and is able to do this e.g. without the necessity of costly remeshing procedures. In addition it might also help to safe computational costs as now background grids can be much coarser. Various FSI-cases of rising complexity conclude the work. △ Less

Submitted 1 August, 2018; originally announced August 2018.

Comments: 32 pages, 16 figures

Monolithic cut finite element based approaches for fluid-structure interaction

Authors: Benedikt Schott, Christoph Ager, Wolfgang A. Wall

Abstract: Cut finite element method (CutFEM) based approaches towards challenging fluid-structure interaction (FSI) are proposed. The different considered methods combine the advantages of competing novel Eulerian (fixed-grid) and established Arbitrary-Lagrangian-Eulerian (ALE) (moving mesh) finite element formulations for the fluid. The objective is to highlight the benefit of using cut finite element tech… ▽ More Cut finite element method (CutFEM) based approaches towards challenging fluid-structure interaction (FSI) are proposed. The different considered methods combine the advantages of competing novel Eulerian (fixed-grid) and established Arbitrary-Lagrangian-Eulerian (ALE) (moving mesh) finite element formulations for the fluid. The objective is to highlight the benefit of using cut finite element techniques for moving domain problems and to demonstrate their high potential with regards to simplified mesh generation, treatment of large structural motions in surrounding flows, capturing boundary layers, their ability to deal with topological changes in the fluid phase and their general straightforward extensibility to other coupled multiphysics problems. In addition to a pure fixed-grid FSI method, also advanced fluid domain decomposition techniques are considered rendering in highly flexible discretization methods for the FSI problem. All stabilized formulations include Nitsche-based weak coupling of the phases supported by the ghost penalty technique for the flow field. For the resulting systems, monolithic solution strategies are presented. Various 2D and 3D FSI-cases of different complexity validate the methods and demonstrate their capabilities and limitations in different situations. △ Less

Submitted 30 July, 2018; originally announced July 2018.

Comments: 41 pages, 16 figures

The MATLAB Toolbox SciXMiner: User's Manual and Programmer's Guide

Authors: Ralf Mikut, Andreas Bartschat, Wolfgang Doneit, Jorge Ángel González Ordiano, Benjamin Schott, Johannes Stegmaier, Simon Waczowicz, Markus Reischl

Abstract: The Matlab toolbox SciXMiner is designed for the visualization and analysis of time series and features with a special focus to classification problems. It was developed at the Institute of Applied Computer Science of the Karlsruhe Institute of Technology (KIT), a member of the Helmholtz Association of German Research Centres in Germany. The aim was to provide an open platform for the development… ▽ More The Matlab toolbox SciXMiner is designed for the visualization and analysis of time series and features with a special focus to classification problems. It was developed at the Institute of Applied Computer Science of the Karlsruhe Institute of Technology (KIT), a member of the Helmholtz Association of German Research Centres in Germany. The aim was to provide an open platform for the development and improvement of data mining methods and its applications to various medical and technical problems. SciXMiner bases on Matlab (tested for the version 2017a). Many functions do not require additional standard toolboxes but some parts of Signal, Statistics and Wavelet toolboxes are used for special cases. The decision to a Matlab-based solution was made to use the wide mathematical functionality of this package provided by The Mathworks Inc. SciXMiner is controlled by a graphical user interface (GUI) with menu items and control elements like popup lists, checkboxes and edit elements. This makes it easier to work with SciXMiner for inexperienced users. Furthermore, an automatization and batch standardization of analyzes is possible using macros. The standard Matlab style using the command line is also available. SciXMiner is an open source software. The download page is http://sourceforge.net/projects/SciXMiner. It is licensed under the conditions of the GNU General Public License (GNU-GPL) of The Free Software Foundation. △ Less

Submitted 11 April, 2017; originally announced April 2017.

Generating Semi-Synthetic Validation Benchmarks for Embryomics

Authors: Johannes Stegmaier, Julian Arz, Benjamin Schott, Jens C. Otte, Andrei Kobitski, G. Ulrich Nienhaus, Uwe Strähle, Peter Sanders, Ralf Mikut

Abstract: Systematic validation is an essential part of algorithm development. The enormous dataset sizes and the complexity observed in many recent time-resolved 3D fluorescence microscopy imaging experiments, however, prohibit a comprehensive manual ground truth generation. Moreover, existing simulated benchmarks in this field are often too simple or too specialized to sufficiently validate the observed i… ▽ More Systematic validation is an essential part of algorithm development. The enormous dataset sizes and the complexity observed in many recent time-resolved 3D fluorescence microscopy imaging experiments, however, prohibit a comprehensive manual ground truth generation. Moreover, existing simulated benchmarks in this field are often too simple or too specialized to sufficiently validate the observed image analysis problems. We present a new semi-synthetic approach to generate realistic 3D+t benchmarks that combines challenging cellular movement dynamics of real embryos with simulated fluorescent nuclei and artificial image distortions including various parametrizable options like cell numbers, acquisition deficiencies or multiview simulations. We successfully applied the approach to simulate the development of a zebrafish embryo with thousands of cells over 14 hours of its early existence. △ Less

Submitted 17 April, 2016; originally announced April 2016.

Comments: Accepted publication at IEEE International Symposium on Biomedical Imaging: From Nano to Macro (ISBI), 2016

Challenges of Integrating A Priori Information Efficiently in the Discovery of Spatio-Temporal Objects in Large Databases

Authors: Benjamin Schott, Johannes Stegmaier, Masanari Takamiya, Ralf Mikut

Abstract: Using the knowledge discovery framework, it is possible to explore object databases and extract groups of objects with highly heterogeneous movement behavior by efficiently integrating a priori knowledge through interacting with the framework. The whole process is modular expandable and is therefore adaptive to any problem formulation. Further, the flexible use of different information allocation… ▽ More Using the knowledge discovery framework, it is possible to explore object databases and extract groups of objects with highly heterogeneous movement behavior by efficiently integrating a priori knowledge through interacting with the framework. The whole process is modular expandable and is therefore adaptive to any problem formulation. Further, the flexible use of different information allocation processes reveal a great potential to efficiently incorporate the a priori knowledge of different users in different ways. Therefore, the stepwise knowledge discovery process embedded in the knowledge discovery framework is described in detail to point out the flexibility of such a system incorporating object databases from different applications. The described framework can be used to gain knowledge out of object databases in many different fields. This knowledge can be used to gain further insights and improve the understanding of underlying phenomena. The functionality of the proposed framework is exemplarily demonstrated using a benchmark database based on real biological object data. △ Less

Submitted 9 February, 2016; originally announced February 2016.

Comments: Proc., 25. Workshop Computational Intelligence, Dortmund, 2015

MSC Class: 92-08; 92C37; 92C55; 68U10