From alexis.girault at kitware.com Fri Jun 22 13:45:52 2018 From: alexis.girault at kitware.com (Alexis Girault) Date: Fri, 22 Jun 2018 13:45:52 -0400 Subject: [Nirfast] NIRFASTSLicer_2.0_major_bugs_report+solutions In-Reply-To: <005c01d3ebd1$24583890$6d08a9b0$@cs.bham.ac.uk> References: <001301d3e93e$0dd5b5b0$29812110$@Domain> <005c01d3ebd1$24583890$6d08a9b0$@cs.bham.ac.uk> Message-ID: Dear Stanislaw, First of all sorry for the late response, I came back from an extended break. Thank you very much for your investigative work and the solutions you provided. Handling different orientations with Slicer has always been tricky, and I implemented most of this as a first-year intern while probably lacking an understanding of the overall issue, so I am really glad that you were able to bring some additional information to solve that issue. We hope to build a community around NIRFAST, and getting such valuable feedback is a great example of how we can benefit from it, so thanks again! #1, #2, #3 I currently have little time allocated to working on NIRFAST, however, I'd be happy to review those changes if you were able to include your fixes in merge requests on github [1][2]. It seems that your bullet points could define separate commits since you have already nicely organized the changes, and the explanations you provided would be great as commit messages/merge request comments. Is that something you would feel comfortable with? If not, would you be able to format your comments and solutions in the issue tracker? [3][4] [1] https://github.com/nirfast-admin/NIRFASTSlicer/pulls [2] https://github.com/nirfast-admin/NIRFAST/pulls [3] https://github.com/nirfast-admin/NIRFASTSlicer/issues [4] https://github.com/nirfast-admin/NIRFAST/issues #4 As soon as we have this merged, we can probably release this as a minor fix version and create then upload packages on github and nirfast.org. #5 Could you specify what you mean by inactive parameters (maybe in the issue tracker)? This module is a Matlab module, and it is needed to use OpenIGTLink and communicate with the Matlab server, but this module is static as far as inputs and UI goes. The new Create Mesh module is a python module that allows having a more dynamic UI and error handling, and it calls the Image2Mesh module internally. We can not remove that module for that reason, but there might be a way to hide it from the modules list: feel free to add an issue in the issue tracker if you believe that is needed. Also, we already have an issue open to support label maps directly as an input in the Create Mesh module: https://github.com/nirfast-admin/NIRFASTSlicer/issues/21 Thanks again, looking forward to hearing more from you. Best, Alexis Girault R&D Engineer in Medical Computing Kitware, Inc. http://www.kitware.com (919) 969-6990 x325 On Mon, May 14, 2018 at 6:16 PM Stanislaw Wojtkiewicz < s.wojtkiewicz at cs.bham.ac.uk> wrote: > Hi All > > > > I want to report bugs in NIRFASTSlicer_v.2.0 mesh generation module. > > I also deliver solutions (attached). > > > > Major problems: > > - translation of coordinates from patient space to x-y-z space -> > crushes the module > > - unfinished implementation of fiducial coordinates processing > (since NIRFASTSlicer v1.0) -> fiducials useless > > > > The ?Create Mesh? crush and fiducials problems can be reproduced using my > head model (generated from MRI with SPM and saved with NIRFASTSlicer2.0). > > (Please mail me for the segmentation ?*.nrrd? file. It?s only 632KB. > However, Nirfast mailing list email size limit is 300KB.) > > The command: > > cd('C:/Program Files/NIRFASTSlicer > 2.0.0/lib/NIRFASTSlicer-2.0/matlab-modules'); > > > Image2Mesh(cli_argsread({'--nirfastDir','D:/PROGRAMS/nirfast8/toolbox','--labelmap','C:/Users/wojtkies/AppData/Local/Temp/NIRFASTSlicer/EICE_vtkMRMLLabelMapVolumeNodeC.nrrd','--fiducials','34.7404,71.4787,25.567','--fiducials','0.909979,77.0329,25.567','--fiducials','-29.8908,69.9639,25.567','--meshdir','D:/DOCUMENTS/GRANTY/BITMAP/Warsaw_meeting_05.2018/TEST/mesh_test','--meshname','nirfast_mesh','--meshtype','Standard','--cell_size','3.5','--cellradiusedge','3','--facet_size','3.5','--facetangle','25','--facetdistance','3'})); > > > > (Win 10, Matlab R2016a, NIRFASTSLicer2.0 ? pure as download, NIRFAST9.0 ? > pure as download) > > > > DETAILS of the two major problems + other minor/major bugs repairs: > > > > #1 > > Coordinates translation > > The current implementation (Image2Mesh.m) assumes only one possible > orientation of object in the patient space. All images are processed as > oriented 'LPS' . The MRI structure I use was saved in > 'right-anterior-superior' (?RAS?) orientation. > > My spacedir matrix (rotation/scale matrix from patient to xyz space) is > as follows (I have 1mm voxels spacing, thus ones in the matrix): > > 0 0 1 > > -1 0 0 > > 0 1 0 > > However, the code assumes nonzero diagonal values (as it is true for ?LPS? > patient space orientations). Thus, the following lines: > > % param.TransformMatrix(1,1) = param.TransformMatrix(1,1) / > abs(spacedir(1,1)); > > % param.TransformMatrix(2,2) = param.TransformMatrix(2,2) / > abs(spacedir(2,2)); > > % param.TransformMatrix(3,3) = param.TransformMatrix(3,3) / > abs(spacedir(3,3)); > > produce NaNs at the diagonal of the transformation matrix. We divide 0 by > 0. This leads to a big crush. > > > > I think, that the best solution is to use affine transform (param.TransformMatrix > = img.ijkToLpsTransform;). This affine matrix is called ?ijkToLpsTransform? > but Slicer puts there correct values no matter what is the patient space > orientation (i.e. it does not have to be LPS). > > The rotation and translation of the model are handled separately despite a > ready to use affine matrix. Is there any reason to not trust this matrix? I > checked it for different DICOMS (a neck and heads carried out by different > hospitals) and this affine transform matrix was always good (for different > patient orientations!). I am not sure if I fully understand the terminology > LPS etc. and what exactly is passed from Slicer (I just tried to reverse > engineer this problem). However, it looks like the proposed solution works. > > > > Briefly, my changes to the Image2Mesh.m file: > > - the affine transform instead of rotation and translation > > The rotation was carried out inside ?Image2Mesh.m? file, the translation > was carried out inside ?RunCGALMeshGenerator.m? file. The affine transform > does all at once. I have moved the affine operation into ?Image2Mesh.m? as > this file is closely related to the Slicer which generates the affine > transformation matrix. And it might be more clear to keep the > transformation as close as possible to the transformation matrix. > > - with the affine transform, there is no need of the > hard/hand-coded repairs of the rotation matrix (is this to match the > fiducials coordinates? I do not understand this rotations around 0X and 0Y > axes.): > > % param.TransformMatrix(1,1) = -1 * param.TransformMatrix(1,1); > > % param.TransformMatrix(2,2) = -1 * param.TransformMatrix(2,2); > > - removed redundancies > > - code polishing, etc. > > > > The fix requires changes in two files: RunCGALMeshGenerator.m (NIRFAST > package) and Image2Mesh.m (Slicer package). Fixed files attached. > > > > #2 > > Unfinished fiducials business > > Fiducials coordinates are translated internally by the Slicer (by the same > vector as in the affine transform in #1). However, they are not rotated. > > Fiducials look like they are always in the same patient space, no matter > what is the data patient orientation. Thus, to convert the fiducials to the > x-y-z coordinates a following rotation matrix should be used: > > -1 0 0 > > 0 -1 0 > > 0 0 1 > > > > I have added the following to the ?Image2Mesh.m? file: > > sdcoords = ([-1 0 0;0 -1 0; 0 0 1]*sdcoords')'; > > This and affine transform of the model brings the model and fiducials to > the same space. > > > > BTW: The fiducials coordinates passed to the ?Image2Mesh.m? function and > saved as a *.csv file are different?. > > > > #3 > > Reading fiducials fails if there is only one fiducial. > > sdcoords = cell2mat(fiducials)'; > > For one fiducial, the ?fiducials? is not a cell and the command fails > returning error. > > > > Problem solved in the attached file. > > > > #4 > > Can you repair/swap RunCGALMeshGenerator.m and Image2Mesh.m files in the > packages to download? > > I am going to run a training session ?from MRI to FEM model? for PhD > students soon (in 2 weeks). The very first thing I will have to do is to > ask everybody to repair tools by replacing files here and there (or use my > Matlab meshing implementation directly on SPM files, which I developed to > test the errors). > > I have to work on DICOM files that confuses NIRFASTSlicer. So, I have no > choice but distribute NIRFASTSlicer with my repaired files. > > > > #5 > > ?Image2Mesh (Matlab)? module? A testing leftover? It has inactive > settings/parameters. These parameters are loaded from the ?Create Mesh? > module. There is no easy way to guess that. I propose to delete this > redundant module. Or repair the inactive options and leave it as backward > compatibility for meshing ?LabelMaps? not ?Segmentations? as in the new > ?Create Mesh? module. Or add option of meshing a ?LabelMap? to the ?Create > Mesh? module? > > > > > > Regards > > Stanislaw Wojtkiewicz, PhD > > > > --- > > School of Computer Science > > University of Birmingham > > B15 2TT > > > > > > > Virus-free. > www.avast.com > > <#m_-9174478099627014613_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2> > _______________________________________________ > Nirfast mailing list > Nirfast at public.kitware.com > https://public.kitware.com/mailman/listinfo/nirfast > -------------- next part -------------- An HTML attachment was scrubbed... URL: From alexis.girault at kitware.com Fri Jun 22 13:49:46 2018 From: alexis.girault at kitware.com (Alexis Girault) Date: Fri, 22 Jun 2018 13:49:46 -0400 Subject: [Nirfast] Failing while trying to graphically place sources/detectors In-Reply-To: <001301d3e7de$87ea1fa0$97be5ee0$@cs.bham.ac.uk> References: <001301d3e7de$87ea1fa0$97be5ee0$@cs.bham.ac.uk> Message-ID: Thank you, Karthik, Stanislaw, I believe Scott and Hamid are aware of limitations with recent matlab versions. Would you be able to add the encountered issue in the project issue tracker? This will allow us to plan the needed work for the next release: https://github.com/nirfast-admin/NIRFAST/issues Best, Alexis Girault R&D Engineer in Medical Computing Kitware, Inc. http://www.kitware.com (919) 969-6990 x325 On Wed, May 9, 2018 at 6:10 PM Stanislaw Wojtkiewicz < s.wojtkiewicz at cs.bham.ac.uk> wrote: > Hi Karthik > > > > I can confirm, the newest Matlab does not like some functions used in > NIRFAST. > > > > The very last issue: > > ?Error using matlab.graphics.axis.Axes/get > There is no x_RenderTransform property on the Axes class.? > > MathWorks removed the ?x_RenderTransform? axes property starting with > Matlab R2014a. More community comments about that here: > https://undocumentedmatlab.com/blog/undocumented-view-transformation-matrix > > > > No easy fix of that (the GUI of source/detector placing might require some > refurbishment). > > > > However, I might have a solution for you. > > > > Here is how to add source/detector manually (using Matlab commands). Step > by step: > > > > #0 > > Do not place points using a mouse. > > > > #1 > > @ the left side of the GUI where you have empty lists of sources and > detectors: put some values in both. E.g. > > 1 2 3 > > 2 4 5 > > For (x y z) coordinates of e.g. 2 sources. Do similar with detectors. > > The 'points placing' with a mouse just reads the (x y z) coordinates from > the figure axes and puts them into the lists. You might try to put the > exact coordinates if you know them. > > #2 > > Save the mesh from the GUI. Step #1 is not necessary. You should be able > to save mesh without fiducial points. However, you will have more work > later :) > > #3 > > Load the mesh in Matlab command window, script, etc.. using the > ?load_mesh? function (mesh = load_mesh(?path_to_mesh/mesh_name?);) > > #4 > > Adjust your sources and detector. > > The mesh structure should have following fields: > > source: [1?1 struct] > > meas: [1?1 struct] > > > > e.g. > > >> mesh.source > > > > ans = > > > > struct with fields: > > > > distributed: 0 > > fixed: 0 > > num: 1 > > coord: [47.0244 59.5354 48.0170] > > fwhm: 0 > > > > and > > > > >> mesh.meas > > > > ans = > > > > struct with fields: > > > > fixed: 0 > > num: [2?1 double] > > coord: [2?3 double] > > int_func: [2?5 double] > > > > Look for the fields meaning on the online documentation. Focus on: > coordinates (coord) and numbering (num). > > > > #5 > > Change the mesh.link field accordingly (using your sources and detectors > numbering). The ?link? field is explained somewhere online. It is basically > a list of source and detector numbers and logical indicator (0, 1) if this > pair is considered as active (if the boundary fluence rate should be > calculated for that pair). > > > > #6 > > Save the mesh using the command ?save_mesh(mesh, ?path_2_mesh/mesh_name?)? > > > > #7 > > Enjoy the mesh. > > The source/detector coordinates do not have to be exactly on the mesh > surface. Try to put them close to the surface. The load_mesh function will > take care of proper positioning if you leave the ?fixed? field set to ?0?. > > > > I hope I helped. > > > > > > > > Regards > > Stanislaw Wojtkiewicz, PhD > > > > --- > > School of Computer Science > > University of Birmingham > > B15 2TT > > > > *From:* Vishwanath, Karthik > *Sent:* 09 May 2018 19:15 > *To:* nirfast at public.kitware.com > *Subject:* [Nirfast] Failing while trying to graphically place > sources/detectors > > > > Hi All, > > Was having trouble creating and placing sources/detectors in a standard > slab mesh using NIRFAST-9.0 on a linux system (Ubuntu 16.04.4 LTS; > 4.4.0-122-generic x86_64 kernel). My MATLAB version is R2017a. > > I've listed sequence of steps with the messages in the workspace window > shown and the fixes I attempted (in blue text) > > ================================================== > > >> nirfast > > Selected menu in GUI: Mesh->Create Mesh ->Simple Shapes -> Standard > > Selected slab and set L/W/H to 20mm and click Done > > Error using textscan > Unknown parameter 'bufsize'. > > Error in read_nod_elm (line 45) > data = textscan(fid,'%u32 %f %f %f%*[^\n]','bufsize',409500); > > Error in checkerboard3d_mm (line 109) > [telem tnode] = read_nod_elm(fnprefix,1); > > Error in make_slab (line 89) > mesh = checkerboard3d_mm(fullfile(outputdir,'test_node_ele.ele'),... > > > Error in create_mesh (line 25) > eval(['mesh = make_' lower(shape) '(sizevar);']); > > Error in gui_create_mesh>done_Callback (line 532) > mesh = evalin('base',content{end}); > > Error in gui_mainfcn (line 95) > feval(varargin{:}); > > Error in gui_create_mesh (line 42) > gui_mainfcn(gui_State, varargin{:}); > > Error in > > matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)gui_create_mesh('done_Callback',hObject,eventdata,guidata(hObject)) > Error while evaluating UIControl Callback. > > > > ================================================== > > Changed line 45 in "toolbox/meshing/tools/read_nod_elm.m" to: data = > textscan(fid,'%u32 %f %f %f%*[^\n]') > > > Selected menu in GUI: Mesh->Create Mesh ->Simple Shapes -> Standard > > Selected slab and set L/W/H to 20mm and click Done > > > > Added Shield: 0.0010 s > Delaunay Triangulation Time: 0.7141 s > Connectivity Time: 0.0359 s > Circumcenters Time: 0.0129 s > 47 th level was reached > 100.0000 % of Tetraedroms were checked > Walking Time: 0.0399 s > Manifold extraction Time: 0.4239 s > Total Time: 1.2299 s > Writing data to file...Done writing mesh to: > Path: /home/karthik/Research/Matlab/lib/nirfast > Filename: test_node_ele.node/.ele > > > --> Beginning mesh generation process, please wait... > Done with sub-volume separation. > =========================================== > Calculating desired length at boundary nodes.. done. > Calculating prism normals and bounding boxes.. done. > Sealing boundary buffer zone.. done. > > Tagging interior nodes... done > -----> Running BSP tree to filter out nodes. > > -----> done. > Writing data to file.Done writing the nodes to: > Path: /tmp > Filename: input4delaunay.a.node > > Writing data to file... Done writing surface to a poly file: > Path: /tmp > Filename: input4delaunay.poly > > ---------> Running Delaunay, please wait.../bin/bash: > /home/karthik/Research/Matlab/lib/nirfast/NIRFAST-9.0/toolbox/meshing/bin/delaunaygen-linux64.exe: > Permission denied > Warning: Delaunay Generator failed. Trying again... > > In checkerboard3d (line 181) > In checkerboard3d_mm (line 167) > In make_slab (line 89) > In create_mesh (line 25) > In gui_create_mesh>done_Callback (line 532) > In gui_mainfcn (line 95) > In gui_create_mesh (line 42) > In > matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)gui_create_mesh('done_Callback',hObject,eventdata,guidata(hObject)) > > /bin/bash: > /home/karthik/Research/Matlab/lib/nirfast/NIRFAST-9.0/toolbox/meshing/bin/delaunaygen-linux64.exe: > Permission denied > Error using checkerboard3d (line 186) > Delaunay Generator failed again. Check your input setting! > > Error in checkerboard3d_mm (line 167) > [mesh.elements, mesh.nodes] = checkerboard3d(telem(:,1:3),tnode,myargs); > > Error in make_slab (line 89) > mesh = checkerboard3d_mm(fullfile(outputdir,'test_node_ele.ele'),... > > > Error in create_mesh (line 25) > eval(['mesh = make_' lower(shape) '(sizevar);']); > > Error in gui_create_mesh>done_Callback (line 532) > mesh = evalin('base',content{end}); > > Error in gui_mainfcn (line 95) > feval(varargin{:}); > > Error in gui_create_mesh (line 42) > gui_mainfcn(gui_State, varargin{:}); > > Error in > matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)gui_create_mesh('done_Callback',hObject,eventdata,guidata(hObject)) > > Error while evaluating UIControl Callback. > > ================================================== > > In shell: > > $ chmod +x > /home/karthik/Research/Matlab/lib/nirfast/NIRFAST-9.0/toolbox/meshing/bin/delaunaygen-linux64.exe > > > > Selected menu in GUI: Mesh->Create Mesh ->Simple Shapes -> Standard > > Selected slab and set L/W/H to 20mm and click Done > > > > Added Shield: 0.0003 s > Delaunay Triangulation Time: 0.4179 s > Connectivity Time: 0.0214 s > Circumcenters Time: 0.0032 s > Warning: Brute continuation necessary > > In MyRobustCrust>Marking (line 408) > In MyRobustCrust (line 134) > In make_slab (line 85) > In create_mesh (line 25) > In gui_create_mesh>done_Callback (line 532) > In gui_mainfcn (line 95) > In gui_create_mesh (line 42) > In > matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)gui_create_mesh('done_Callback',hObject,eventdata,guidata(hObject)) > > Warning: 1000 th level was reached\n > > In MyRobustCrust>Marking (line 440) > In MyRobustCrust (line 134) > In make_slab (line 85) > In create_mesh (line 25) > In gui_create_mesh>done_Callback (line 532) > In gui_mainfcn (line 95) > In gui_create_mesh (line 42) > In > matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)gui_create_mesh('done_Callback',hObject,eventdata,guidata(hObject)) > > 99.7824 % of Tetraedroms were checked > Walking Time: 0.0985 s > Manifold extraction Time: 0.2470 s > Total Time: 0.7971 s > Writing data to file...Done writing mesh to: > Path: /home/karthik/Research/Matlab/lib/nirfast > Filename: test_node_ele.node/.ele > > > --> Beginning mesh generation process, please wait... > Done with sub-volume separation. > =========================================== > Calculating desired length at boundary nodes.. done. > Calculating prism normals and bounding boxes.. done. > Sealing boundary buffer zone.. done. > > Tagging interior nodes... done > -----> Running BSP tree to filter out nodes. > > -----> done. > Writing data to file.Done writing the nodes to: > Path: /tmp > Filename: input4delaunay.a.node > > Writing data to file... Done writing surface to a poly file: > Path: /tmp > Filename: input4delaunay.poly > > ---------> Running Delaunay, please wait... done. <--------- > > > > --> Finished mesh generation. > Undefined function or variable 'myflag'. > > Error in optimize_mesh_gui>optimize_mesh_gui_OpeningFcn (line 78) > if myflag > > Error in gui_mainfcn (line 220) > feval(gui_State.gui_OpeningFcn, gui_hFigure, [], guidata(gui_hFigure), > varargin{:}); > > Error in optimize_mesh_gui (line 40) > [varargout{1:nargout}] = gui_mainfcn(gui_State, varargin{:}); > > Error in gui_create_mesh>done_Callback > > Error in gui_mainfcn (line 95) > feval(varargin{:}); > > Error in gui_create_mesh (line 42) > gui_mainfcn(gui_State, varargin{:}); > > Error in > matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)gui_create_mesh('done_Callback',hObject,eventdata,guidata(hObject)) > > Error while evaluating UIControl Callback. > > > > ================================================== > > Added line 64 in toolbox/meshing/tools/optimize_mesh_gui.m: > > 64 myflag=''; > 65 if(nargin > 3) > > Selected menu in GUI: Mesh->Create Mesh ->Simple Shapes -> Standard > > Selected slab and set L/W/H to 20mm and click Done > > > > The mesh is generated and the figure dialog for "Place sources and > detectors" comes up. > > The mesh is viewable, can be rotated and panned using the tools in the > figure window. However, clicking on the mesh generates the following error: > > Error using matlab.graphics.axis.Axes/get > There is no x_RenderTransform property on the Axes class. > > Error in select3d>local_Data2PixelTransform (line 361) > xform = get(ax,'x_RenderTransform'); > > > Error in select3d (line 208) > xvert = local_Data2PixelTransform(ax,vert)'; > > > Error in gui_place_sources_detectors>figure1_WindowButtonDownFcn (line 386) > p = select3d; > > Error in gui_mainfcn (line 95) > feval(varargin{:}); > > Error in gui_place_sources_detectors (line 42) > gui_mainfcn(gui_State, varargin{:}); > > Error in > matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)gui_place_sources_detectors('figure1_WindowButtonDownFcn',hObject,eventdata,guidata(hObject)) > > Error while evaluating Figure WindowButtonDownFcn. > > > ================================================== > > > > Was not sure how to proceed from there. Let me know if there is any other > information needed to help track/fix this. > > Thanks, > > -- > Karthik Vishwanath, PhD > James C. & Carole E. Garland Professorship > > Office: 109 Kreger Hall > > Dept. of Physics, Miami University, Oxford OH > > Tel: (513) 529-2315 > > Web: OSIM Lab > > > Virus-free. > www.avast.com > > <#m_-7383975974148153812_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2> > _______________________________________________ > Nirfast mailing list > Nirfast at public.kitware.com > https://public.kitware.com/mailman/listinfo/nirfast > -------------- next part -------------- An HTML attachment was scrubbed... URL: From alexis.girault at kitware.com Fri Jun 22 13:52:28 2018 From: alexis.girault at kitware.com (Alexis Girault) Date: Fri, 22 Jun 2018 13:52:28 -0400 Subject: [Nirfast] Mathworks removed 'cholinc' function In-Reply-To: <001b01d3ec2d$6dbe2e20$493a8a60$@cs.bham.ac.uk> References: <001b01d3ec2d$6dbe2e20$493a8a60$@cs.bham.ac.uk> Message-ID: Hamid, Scott, Same as previously: looks like a NIRFAST-Matlab question. Would you like to have this filed in our NIRFAST issue tracker? https://github.com/nirfast-admin/NIRFAST/issues Best, Alexis Girault R&D Engineer in Medical Computing Kitware, Inc. http://www.kitware.com (919) 969-6990 x325 On Tue, May 15, 2018 at 5:16 AM Stanislaw Wojtkiewicz < s.wojtkiewicz at cs.bham.ac.uk> wrote: > Hi All > > > > One more problem to report. > > In time-resolved modality (file ?femdata_stnd_tr.m?) the incomplete > Cholesky factorisation is used to generate a preconditioner: > > % R = cholinc(MASS1,1e-3); > > However, Mathworks removed this function completely since it was not a > true incomplete Cholesky (it utilized incomplete LU with pivoting). > > I suggest to replace the missing function by: > > R = ichol(MASS1,struct('type','ict','droptol',1e-3,'shape','upper')); > > Works in my case. > > > > Regards > > Stanislaw Wojtkiewicz, PhD > > > > --- > > School of Computer Science > > University of Birmingham > > B15 2TT > > > > > Virus-free. > www.avast.com > > <#m_-1140050059574826654_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2> > _______________________________________________ > Nirfast mailing list > Nirfast at public.kitware.com > https://public.kitware.com/mailman/listinfo/nirfast > -------------- next part -------------- An HTML attachment was scrubbed... URL: