eidors 3 - carleton · eidorsversion 3.8 andy adler1, alistair boyle1, michael g. crabb2, hervÉ...

1
EIDORS V ERSION 3.8 A NDY A DLER 1 ,A LISTAIR B OYLE 1 ,M ICHAEL G. C RABB 2 ,H ERVÉ G AGNON 1 , B ARTLOMIEJ G RYCHTOL 3 ,N OLWENN L ESPARRE 4 ,W ILLIAM R. B. L IONHEART 2 1 Carleton University, Ottawa, Canada 2 University of Manchester, Manchester, UK 3 Fraunhofer Project Group for Automation in Medicine and Biotechnology PAMB, Mannheim, Germany 4 IRSN, B.P. 17, 92262 Fontenay-aux-Roses Cedex, France N EW R ELEASE We are pleased to announce the release of EIDORS 3.8 [1]. The software is available at www.eidors.org licensed under the GNU GPLv2 (or GPLv3). EIDORS 3.8 version EIDORS aims to provide free software al- gorithms for forward modelling and inverse solutions of Electrical Impedance and (to some extent) Diffusion-based Optical Tomo- graphy, in medical, industrial and geophys- ical settings and to share data and promote collaboration. N EW F EATURES Release 3.8 of EIDORS builds upon a strong foundation in reconstruction al- gorithms, adding and improving a number of aspects. More stable iterative absolute inverse solvers (both Gauss-Newton and Conjugate- Gradient). Greater flexibility in parametrization choices. Native handling of unit scaling (10 x , e x , ln x, log 10 x), and arbitrary units. Natural limits for σ> 0. GREIT reconstructions in 3D Speed optimizations: improved Jacobian calculation, faster cache handling, and faster forward solutions. Improved interfaces to NetGen and visu- alization. Compound and point electrodes in NetGen. Analytic calculation of dual-mesh inter- polations (coarse to fine) Support for second and third order mesh elements. Support for Dräger and Swisstom file formats Expanded shape library R EFERENCES [1] Adler A, Boyle A, Crabb MG et al, EIDORS v3.8, Zenodo, DOI:10.5281/zenodo.17559, 2015. [2] Vauhkonen M, Lionheart WRB, Heikkinen L et al, Physiol Meas, 22:107–111, 2001. [3] Polydorides N, Image Reconstruction Algorithms for Soft-Field Tomography, Ph.D. thesis, University of Manchester, UK, 2002. [4] Polydorides N, Lionheart WRB, Meas Sci and Tech, 13:1871–1883, 2002. [5] Adler A, Lionheart WRB, Physiol Meas, 27:S25–S42, 2006. S UCCESSES The structure of EIDORS has been relat- ively stable due, in part, to some early design choices: a modular framework and data structure, cross-platform support, integra- tion of meshing, tutorials, and the contrib- uted data repository. These aspects, along with an open source code-base, have enabled EIDORS to maintain research relevance. C HALLENGES A number of challenges inherent in the implementation of EIDORS as a Matlab- based toolkit continue to recur. There is no real Object Oriented framework: no reflec- tion, protection, or automatic management of errors. Versions of Matlab frequently vary in confounding ways that make maintaining a toolkit across multiple Matlab versions dif- ficult. This is particularly prevalent for Win- dows users and “mex" file compilation. The data structure and subfunction complexity in EIDORS are a source of confusion for be- ginners. Despite these challenges, EIDORS contin- ues to develop, grow and expand into new areas. Presenting version 3.8! G ROWTH EIDORS-related citations continue to grow. Current citation results are shown in table 1. The EIDORS code-base is stable with significant effort being applied to improving test cover- age, refining performance and implementing new features (fig. 1). In 2012, a dev staging area was created for contributions in progress. 2004 2006 2008 2010 2012 2014 0 20 40 60 80 Lines of Code (x1000) Total Eidors Tutorials Dev Figure 1: Lines of Code (LoC) in Matlab files in the EIDORS code-base vs. time; Total (red), Eidors (i.e. release branch, olive), Tutorials (green), development code (blue). Releases are indicated by gray bars. Table 1: EIDORS Citations (May 2015, scholar.google.com). Paper Date Citations [2] A MATLAB package for the EIDORS project ... 2001 159 [3] Image reconstruction algorithms for ... 2002 88 [4] A Matlab toolkit for three-dimensional ... 2002 293 [5] Uses and abuses of EIDORS: An extensible ... 2006 184 E XAMPLE FEATURES Ventilation without/with lung injury 3D GREIT reconstruction New Netgen interface Streamlines near electrode vs. z c A CKNOWLEDGEMENTS Recent funding for EIDORS development thanks to: Swisstom AG, NSERC Canada, EPSRC UK, and IRSN France.

Upload: others

Post on 20-Aug-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: EIDORS 3 - Carleton · EIDORSVERSION 3.8 ANDY ADLER1, ALISTAIR BOYLE1, MICHAEL G. CRABB2, HERVÉ GAGNON1, BARTŁOMIEJ GRYCHTOL3, NOLWENN LESPARRE4, WILLIAM R. B. LIONHEART2 1 Carleton

EIDORS VERSION 3.8ANDY ADLER1, ALISTAIR BOYLE1, MICHAEL G. CRABB2, HERVÉ GAGNON1,

BARTŁOMIEJ GRYCHTOL3, NOLWENN LESPARRE4, WILLIAM R. B. LIONHEART2

1Carleton University, Ottawa, Canada 2University of Manchester, Manchester, UK3Fraunhofer Project Group for Automation in Medicine and Biotechnology PAMB, Mannheim, Germany

4IRSN, B.P. 17, 92262 Fontenay-aux-Roses Cedex, France

NEW RELEASE

We are pleased to announce the release ofEIDORS 3.8 [1]. The software is available atwww.eidors.org licensed under the GNUGPLv2 (or GPLv3).

EIDORS

3.8version

EIDORS aims to provide free software al-gorithms for forward modelling and inversesolutions of Electrical Impedance and (tosome extent) Diffusion-based Optical Tomo-graphy, in medical, industrial and geophys-ical settings and to share data and promotecollaboration.

NEW FEATURESRelease 3.8 of EIDORS builds upon

a strong foundation in reconstruction al-gorithms, adding and improving a numberof aspects.

More stable iterative absolute inversesolvers (both Gauss-Newton and Conjugate-Gradient).

Greater flexibility in parametrizationchoices.

Native handling of unit scaling (10x, ex,lnx, log10 x), and arbitrary units. Naturallimits for σ > 0.

GREIT reconstructions in 3D

Speed optimizations: improved Jacobiancalculation, faster cache handling, and fasterforward solutions.

Improved interfaces to NetGen and visu-alization. Compound and point electrodes inNetGen.

Analytic calculation of dual-mesh inter-polations (coarse to fine)

Support for second and third order meshelements.

Support for Dräger and Swisstom fileformats

Expanded shape library

REFERENCES

[1] Adler A, Boyle A, Crabb MG et al, EIDORS v3.8,Zenodo, DOI:10.5281/zenodo.17559, 2015.

[2] Vauhkonen M, Lionheart WRB, Heikkinen L et al,Physiol Meas, 22:107–111, 2001.

[3] Polydorides N, Image Reconstruction Algorithms forSoft-Field Tomography, Ph.D. thesis, University ofManchester, UK, 2002.

[4] Polydorides N, Lionheart WRB, Meas Sci and Tech,13:1871–1883, 2002.

[5] Adler A, Lionheart WRB, Physiol Meas, 27:S25–S42,2006.

SUCCESSESThe structure of EIDORS has been relat-

ively stable due, in part, to some early designchoices: a modular framework and datastructure, cross-platform support, integra-tion of meshing, tutorials, and the contrib-uted data repository. These aspects, alongwith an open source code-base, have enabledEIDORS to maintain research relevance.

CHALLENGESA number of challenges inherent in the

implementation of EIDORS as a Matlab-based toolkit continue to recur. There is noreal Object Oriented framework: no reflec-tion, protection, or automatic managementof errors. Versions of Matlab frequently varyin confounding ways that make maintaininga toolkit across multiple Matlab versions dif-ficult. This is particularly prevalent for Win-dows users and “mex" file compilation. Thedata structure and subfunction complexityin EIDORS are a source of confusion for be-ginners.

Despite these challenges, EIDORS contin-ues to develop, grow and expand into newareas. Presenting version 3.8!

GROWTHEIDORS-related citations continue to grow. Current citation results are shown in table 1.

The EIDORS code-base is stable with significant effort being applied to improving test cover-age, refining performance and implementing new features (fig. 1). In 2012, a dev staging areawas created for contributions in progress.

2004 2006 2008 2010 2012 20140

20

40

60

80

Line

s of

Cod

e (x

1000

)

TotalEidorsTutorialsDev

Figure 1: Lines of Code (LoC) in Matlab files in the EIDORS code-base vs. time;Total (red), Eidors (i.e. release branch, olive), Tutorials (green), developmentcode (blue). Releases are indicated by gray bars.

Table 1: EIDORS Citations (May 2015, scholar.google.com).Paper Date Citations

[2] A MATLAB package for the EIDORS project ... 2001 159[3] Image reconstruction algorithms for ... 2002 88[4] A Matlab toolkit for three-dimensional ... 2002 293[5] Uses and abuses of EIDORS: An extensible ... 2006 184

EXAMPLE FEATURES

Ventilation without/with lung injury

3D GREIT reconstruction

New Netgen interface

Streamlines near electrode vs. zc

ACKNOWLEDGEMENTSRecent funding for EIDORS development thanks to:

Swisstom AG, NSERC Canada, EPSRC UK, and IRSN France.