vigente / gerardus Goto Github PK

Code already written, just needs the help and copyright headers.

No sufolders inside of toolboxes. 

https://code.google.com/p/gerardus/source/browse/#svn%2Ftrunk%2Fmatlab%2FCardiac
Toolbox%2FgetCMR42Contours%253Fstate%253Dclosed

And getCMR42Contours is derivative from a third party toolbox too.

This should go in the ThirdPartyToolbox, in a subfolder with the toolbox name, 
e.g. ThirdPartyToolbox/BlahBlahToolbox.

First, the original third party code is added and commited. 

Then, you apply your changes, and add near their copyright/licence snippet 
something along the lines of 

% This is a derivative work of the original BlahBlah toolbox.
% Author: Tasos <[email protected]>
% Version: 0.1.0
% $Date$
% $Rev$
%
% Summary of changes:
%
% * modified GUI to select several files at the same time.
% * change of colours.

The script add_gerardus_path.m will pick up the new 
ThirdPartyToolbox/BlahBlahToolbox folder and add it to the path, anyway.

private-gerardus:

matlab/CardiacToolbox/local_reference_frame

Decide on one of my functions to commit, ensure header etc in Gerardus format, 
commit it.

Ben to check his functions work with new DICOM read function added by Chris

Possibly create a toolbox for all the DT code that is not cardiac-specific. e.g 
Darryl's dt2image, fit_dt & then Jo's work (yet to be committed).

B-splines in ITK - elastix use an underlying grid, but elastix doesn't return 
the location of the grid points. For illustration purposes, it would be useful 
to have a Matlab function that computes those points.

The details of how the grid is computed are here:

elastix\trunkpublic\src\Common\Transforms\itkGridScheduleComputer

1) Whitespaces missing, function name in help header must be equal to actual 
function name, and end with full stop: 

"%DT_TO_IMAGE Inverses the transform in fit_DT"

to

"% DT_to_image  Inverses the transform in fit_DT."

https://code.google.com/p/gerardus/source/browse/trunk/matlab/CardiacToolbox/DT_
to_image.m#2

2) Isn't it "Inverts" instead of "Inverses"?

3) Matlab tends to call this kind of function "dt2image.m" instead of 
"DT_to_image.m" (e.g. ind2sub, sph2cart, etc.). And generally, we want lower 
case for the beginning of function names.

Create sanity check function for loaded dicom data

surface_param: 

     XY plane:
                  'xy':       Simple projection on XY.
                  'pca':      Projection on main PCA plane.
                  'isomap':   Isomap (with Dijkstra's shortest path).
                  'cmdsmap':  MDS mapping, global neighbourhood, classical
                              MDS.
                  'lmdscale': MDS mapping, local neighbourhood, local MDS.

      Unit sphere:
                  'sphproj':  Simple projection on sphere around centroid.
                  'cald':     CALD method by Shen & Makedon 2006.
                  'smdscale': Spherical extension of lmdscale.


tri_sphparam:

      'cmdscale': Classical Multidimensional Scaling (MDS).

      'smacof':   Unconstrained SMACOF, followed by projection of points on
                  sphere.

      'consmacof-global': Constrained SMACOF with untangling of all
                  vertices simulataneously (too slow except for small
                  problems).

      'consmacof-local': Constrained SMACOF with local untangling of
                  connected vertices. If the algorithm cannot find a way to
                  untangle a local component, it leaves its vertices
                  untouched. This method is parallelized. To take advantage
                  of multiple threads, before running this function:

                  % activate pool of workers
                  matlabpool

                  % select number of parallel threads
                  myCluster = parcluster();
                  myCluster.NumWorkers = 6;

1) Line

% Reformats nD-arrayed 2D image data for display as 2D montage.

should be changed to

% imagesc_nD  Reformats nD-arrayed 2D image data for display as 2D montage.

2) Line "% IM_IN  = nD data where first two dimensions correspond to a 2D 
image. 3rd to nth dimension are arrays of" is too long. This makes "help 
imagesc_nD" difficult to read. Matlab has smart justify to cut lines to a 
maximum length.

I will submit the cmr42 contour reader to gerardus.
Added as an issue to keep me accountable :p

In matlab/CardiacToolbox/local_reference_frame.m:

% [rv, cv, lv, sf] = local_reference_frame(mask, 'first', 50);
% 
% 
% Author: Darryl McClymont <[email protected]>

should be

% [rv, cv, lv, sf] = local_reference_frame(mask, 'first', 50);


% Author: Darryl McClymont <[email protected]>

Otherwise, when you run "help local_reference_frame" in Matlab you get the 
whole licence block.

I think that property may be missing in some of the files, which translates 
into bad EOLs when opening with notepad.exe in Windows

Which image causes the new version reference frame code to crash?

Currently, our CMake build system finds the installed Matlab version by running 
something like "matlab -version". However, this is a bit slow, and requires to 
be connected to the Uni VPN for Matlab installs that depend on the central 
licence server. 

So is there a way of finding out the version without having to run matlab?

We have been discussing how to generate documentation for the functions in the 
project.

It has to be automatic. E.g. we write a doc_foo.m script for function foo.m, 
with examples on how to use it. This doc_foo.m generates some images too. Then, 
we run something like Matlab Publisher (as suggested by Tasos), and this 
produces XML, HTML or Wiki code, that can automatically be commited to 
gerardus/wiki/.

Darryl has added a MEX function

FiltersToolbox/forward_TV_mex.cpp

Ramon to integrate in the CMake build system.

Why isn't there a CMakeLists.txt file in matlab/PointsToolbox?

Great work. Just a small thing. The help header does not contain the syntax(es) 
of the function.

E.g. for function foo(), you would have

% Y = FOO(X)
%
%   X is a scalar between 0 and 1.
%
%   Y is the square value of X.

See for example 
https://code.google.com/p/gerardus/source/browse/trunk/matlab/PointsToolbox/plot
_dmatrix.m

Format from gmsh to Chaste

Insert a spatial check when slices are ordered, such that the slices are 
ordered in space rather then by selection. 

If the m4 library is not installed, CMake will do the configuration step 
("cmake ..") without problem, but then the build ("make") will fail when 
compiling the GMP library. 

If you look into the GMP compilation log, you see that the problem is that m4 
is missing, but this may be a bit esoteric for most users.

The desired behaviour would be that cmake detects that m4 (or any other 
dependency) is missing, and give a clear error at the configuration stage.

This is something that happens at least with the linux build, not sure whether 
this is also a problem in Windows or MacOS X.

It would simplify the scripts that we use for the local coordinate system work 
if they used the scimat format.

We have to meet

Visualization functions for DICOM struct

make any visualization functions consistent with DICOM struct format generated 
for cine

Ask Adam to join

Can be replaced by:
cross(A, B, 4)

Add dicom read tools, available from MATLAB file exchange.
Upload my edits to this code with added functionality.

Missing:
 copyright, licence, subversion properties (Rev, Date and native), checking input/output arguments.

Format problems:
 function names must start with lower case.

matlab/CardiacToolbox/ImgCrossProduct.m
matlab/CardiacToolbox/ImgDotProduct.m
matlab/CardiacToolbox/ImgNormalise.m
matlab/CardiacToolbox/IsImgNormalised.m
matlab/CardiacToolbox/conv_hull_3D.m

Revision: 1659

Good work, Ben, although as a minor issue, the commit has kind of broken the 
function ;). I have left notes directly on the code with the fixes that need to 
be done. Your should be able to see them here:

https://code.google.com/p/gerardus/source/browse/trunk/matlab/FiltersToolbox/sci
mat_crop.m

Please have a look, try to fix it, and let me know if you need help.

Look for functions that can be contributed. 

Create small DICOM dataset that can be used as a test Dataset. ( Load DIMCOM 
and vizualize it). 
If public domain images, the dataset will go in public Gerardus. If images from 
Biomedia Oxford clinical images, then the dataset will go into private 
Gerardus.  

path: matlab/test/data/dcm

When you run the program extractVoxelCoordinatesFromSegmentationMask with
the -i flag, i.e. the output is formatted as voxel indices instead of real
world coordinates, the x- and y-coordinates are swapped.

A 3D volume voxel with indices

index[0], index[1], index[2]

maps to real world coordinates as

f(index[1]), f(index[0]), f(index[2])

That is, you need to swap index[0] and index[1].

Convert mex TV into normal matlab format.

HISTOLOGY_BLOCKFACE_REG  Register deformed histology volume to blockface.

Just do it

TV filters should be added to Gerardus.

when calling the function 

ImOut = itk_imfilter('advess',scimatIn,1,6,6,5);

I get an image with only minor edge effects different to the original. When 
using a pure C++ itk implementation I get a much changed image. 

link to images: https://www.dropbox.com/s/53xxj2vq5iggv47/AdvessImages.zip?dl=0

Code that was mentioned in 6th Gerardus meeting to be committed.

Once it is committed, add solveSOR to CMAKE build system as a mex function

Install TortoiseSVN.

Check out Gerardus to your hard drive.

What steps will reproduce the problem?
1. For example, rotate 90º to one side an image foo.mha with rotate3DImage 

What is the expected output? What do you see instead?

Assume that foo.mha has resolution [sx, sy, sz].

The new image will have resolution [sx, sy, sz] too, but this means that in
fact the image content resolution has changed. 

If the image has been rotated, now the resolution should be e.g. [sy, sx, sz].

ChangeLog forgotten

Is CardiacToolbox the appropriate toolbox?

Isn't this the same as A*x?

shell-scripts/batch_process_subject.sh

Ramon and I had a quick chat about the licencing nature of the project. The 
project is currently licenced under the GPL with the main copyright holder 
stated as the University of Oxford.

I was thinking that perhaps a more substantial discussion would be useful on 
this topic, and it's something that should be addressed relatively early, 
before we commit to substantial amounts of code / contributions. Ramon 
encouraged me to post this as an 'issue' here.

Points to address would be:
* Is the current licence the best for the project
* Does it address the academic nature of the project (for instance, it does not 
force or encourage people to cite the project or the individual author or 
associated papers when using / modifying the code).
* What is the legal situation with the university being the main copyright 
holder? Should the project has its own autonomous copyright? Do individual 
authors / contributors have a say / rights over their work? Does submitting to 
the project 'lock' a contributor out of their rights to that code?
* What is the legal position with using third-party code? For example, if we're 
using a tool which has been modified from matlab's fileexchange, does this 
affect us as a project? (see: 
http://wiki.octave.org/FAQ#Why_can.27t_I_use_code_from_File_Exchange_in_Octave.3
F_It.27s_released_under_a_BSD_license.21 )

I would be keen to have a better idea about these issues before committing 
substantially to the project; particularly the points about citeability and 
code ownership.

Let me know your thoughts :)

Look for a way to merge changes made by Tasos and by Chris to dicomdir code.

To be discussed with Tasos on Tuesday.

Move code in personal script "histology_alone_mouse.m" to a function in the 
repository.

Syntax missing in help header for FORWARD_TV

We have to make papers reproducible. Code for papers gets developed in 
private-gerardus, and when they are published, move to public gerardus.