EIDORS

EIDORS: Electrical Impedance Tomography and Diffuse Optical Tomography Reconstruction Software

EIDORS (mirror)
Main
Documentation
Tutorials
− Image Reconst
− Data Structures
− Applications
− FEM Modelling
− GREIT
− Old tutorials
− Workshop
Download
Contrib Data
GREIT
Browse Docs
Browse SVN

News
Mailing list
(archive)
FAQ
Developer

Hosted by

GREIT Reconstruction in 3D

Code for Reconstruction using GREIT in 3D was recently added. If you have EIDORS v3.9, you will need the following two files:

For 3D distributions, using GREIT 3D is much better, even if you only want a single slice. See, for example

Grychtol et al, Focusing EIT reconstructions using two electrode planes p. 17 Conf. EIT 2017, Dartmouth, NH, USA, June 21−24, 2017.

This code is also very slow to calculate. The actual reconstruction is fast. There are numerous improvements possible, and they're being worked on.

Model of a tank with an object moving vertically

%STEP 1: Simple model with vertical targets
posns= linspace(0.5,3.5,13);
str=''; for i=1:length(posns);
   extra{i} = sprintf('ball%03d',round(posns(i)*100));
   str = [str,sprintf('solid %s=sphere(0.5,0,%f;0.1); ', extra{i}, posns(i))];
end
extra{i+1} = str;
fmdl= ng_mk_cyl_models([4,1,.2],[16,1.5,2.5],[0.05],extra); 
fmdl = mdl_normalize(fmdl, 0);
[~,fmdl] = elec_rearrange([16,2],'square', fmdl);

img= mk_image(fmdl,1);
img.elem_data(vertcat(fmdl.mat_idx{2:end}))= 2;
opt.viewpoint = struct('az',0,'el',10);
clf;show_fem_enhanced(img,opt);

print_convert GREIT3D_tank01a.jpg

Figure: Simulation cylindrical tank with objects in a vertical plane

Simulate data from tank

% STEP 2: Simulate data
[fmdl.stimulation,fmdl.meas_select] = mk_stim_patterns(32,1,[0,5],[0,5],{},1);
img= mk_image(fmdl,1);
vh = fwd_solve(img);
for i=1:length(posns)-4;
   img= mk_image(fmdl,1);
   img.elem_data(fmdl.mat_idx{i+1}) = 2;
   vi{i} = fwd_solve(img);
end;

Create reconstruction model

A new model to ensure we don't commit an inverse crime. Note that we use the "square" electrode configuration with a "skip 4" stimulation and measurement pattern.

% STEP 3: Reconstruction model 
fmdl= ng_mk_cyl_models([4,1,.5],[16,1.5,2.5],[0.05]);
[fmdl.stimulation,fmdl.meas_select] = mk_stim_patterns(32,1,[0,5],[0,5],{},1);
fmdl = mdl_normalize(fmdl, 0);
[~,fmdl] = elec_rearrange([16,2],'square', fmdl);

GREIT 2D

Here we build a 2D GREIT reconstruction, using training targets only in the centre plane.

% STEP 4: Reconstruction using GREIT 2D
   opt.imgsz = [32 32];
   opt.square_pixels = true;
   opt.noise_figure = 0.5;
   img = mk_image(fmdl,1);
   imdl2 = mk_GREIT_model(img, 0.25, [], opt);

And reconstruct the first 9 targets, from the bottom

img = inv_solve(imdl2,vh,[vi{:}]);
img.show_slices.img_cols= 9;
show_slices(img);

print_convert GREIT3D_tank05a.jpg

Figure: Reconstruction of 9 targets moving from the bottom of the tank using a 2D GREIT version

GREIT 3D

Here we build a 3D GREIT reconstruction, using training targets only in all planes

% STEP 4: Reconstruction using GREIT 3D
   vopt.imgsz = [32 32];
   vopt.zvec = linspace( 0.75,3.25,6);
   vopt.save_memory = 1;
   opt.noise_figure = 2.0;
   [imdl,opt.distr] = GREIT3D_distribution(fmdl, vopt);
   imdl3= mk_GREIT_model(imdl, 0.20, [], opt);

And reconstruct the first 9 targets, from the bottom

img = inv_solve(imdl3,vh,[vi{:}]);

img.show_slices.img_cols= 9;
levels = [inf,inf,1.5;
          inf,inf,2.0];
show_slices(img,levels);

print_convert GREIT3D_tank07a.jpg

img.elem_data = img.elem_data(:,4); % choose 4th image
show_fem(img);
print_convert GREIT3D_tank07b.jpg

show_3d_slices(img,[1.5,2],0,0); view(-16,18);
print_convert GREIT3D_tank07c.jpg

Figure: Left Reconstruction of 9 targets moving from the bottom of the tank using a 3D GREIT version, in two planes. The 4th image is then shown as a volume (Centre) or through cut-planes (Right)

GREIT 3D − 2D slices

In many cases, what is wanted are reconstructions to a slices of the volume, but using the GREIT 3D approach. This is what the solve_RM_2Dslice functions does.

imdl3a= solve_RM_2Dslice(imdl3,1.5);

The single-plane reconstruction matrix is then attached to imdl3a.solve_use_matrix.RM.

img = inv_solve(imdl3a,vh,[vi{:}]);
img.show_slices.img_cols= 9;
show_slices(img);

print_convert GREIT3D_tank09a.jpg

Figure: Reconstruction of 9 targets moving from the bottom of the tank using a slice at z=1.5 of the 3D GREIT version

Last Modified: $Date: 2018-03-28 15:01:38 -0400 (Wed, 28 Mar 2018) $ by $Author: aadler $