Index in position 1 exceeds array bounds (must not exceed 64)

I am plotting 4 different thresholded t-maps using .nii files. My first three t-maps are executed correctly. But there seems to be a problem with my fourth. Meaning that using the same script (and functions) works 3 out of 4 times. I attached error below. Matlab R2018a.

function Y_rgb = sd_slice_to_rgb(Y, layer, settings)
% Panel dimensions
pandims             = settings.slice.pandims;

% Use slover to scale slice to color map indices
Y_map_i             = pr_scaletocmap(Y, ...
                                     layer.color.range(1), ...
                                     layer.color.range(2), ...
                                     layer.color.map, ...
                                     [0 256 0]);

% Convert to RGB - Error in next line
Y_rgb                       = reshape(layer.color.map(Y_map_i(:),:), ...
                                      pandims);
Y_rgb(Y_rgb > 1)            = 1;  
end

% sd_display (line 93)
Y_rgb = sd_slice_to_rgb(Y_c,layers(i_layer),settings);

[Walter Roberson]

Some element of Y_map_i is greater than 64 but layer.color.map only has 64 elements.

We do not know anything about pr_scaletocmap to know how that could come about.

[Janez Novak]

function [img, badvals]=pr_scaletocmap(inpimg,mn,mx,cmap,lrn)
% inpimg     - matrix containing image to scale
% mn         - image value that maps to first value of colormap
% mx         - image value that maps to last value of colormap
% cmap       - 3xN colormap
% lrn        - 1x3 vector,
%              If lrn value is 0, then colormap values are set to 1, and
%              indices to these values are returned in badvals (below)
% img        - inpimg scaled between 1 and (size(cmap, 1))
% badvals    - indices into inpimg containing values out of range, as
cml = size(cmap,1);

if nargin < 5
  lrn = [1 cml 0];
end

img = (inpimg-mn)/(mx-mn);  % img normalized to mn=0,mx=1
if cml==1 % values between 0 and 1 -> 1
  img(img>=0 & img<=1)=1;
else
  img = img*(cml-1)+1;
end
outvals = {img<1, img>cml, isnan(img)};
img= round(img);
badvals = zeros(size(img));
for i = 1:length(lrn)
  if lrn(i)
    img(outvals{i}) = lrn(i);
  else
    badvals = badvals | outvals{i};
    img(outvals{i}) = 1;
  end    

Answer 1

[Walter Roberson]

                                     [0 256 0]);

The 256 there instructs that any values that would map outside of the colormap length should be replaced with 256, no matter what the length of the colormap is. You have at least one value that maps to outside the colormap, such as if your layer.color.range values are not correct for Y

[Janez Novak]

Using function slover inside SPM module, one can plot threshold t-map with GUI.

slover('basic_ui');

If I apply the same values for layer.color.range as in my script, I get correct plot. One threshold t-map goes from [1.67 5.8834] and the other one from [1.67 10.8384].

So I can't fully grasp how to get rid of my error.

[Walter Roberson]

Put a breakpoint at

outvals = {img<1, img>cml, isnan(img)};

and execute the code. When it gets to the breakpoint, step one line so that outvals is assigned to. Now ask

nnz(outvals{2})

if the reported number is not zero then

    img(outvals{i}) = lrn(i);

is going to have an meaningful effect.

When that happens, you need to examine

  max(inpimg(:))
  min(inpimg(:))
  mx
  mn
  max(img(:))
  cml
  [BR, BC] = find(inpimg > mx, 1);
  if ~isempty(BR)
      fprintf('inpimg(%d,%d) = %g which is greater than the declared max, %g\n', BR, BC, inpimg(BR,BC), mx)
  end