Homec4science
Diffusion ShoulderDB_ImportSegmentMeasureAnalyze (aterrier-plotUpdate) Browse

R8218/private/dicom_load_dictionary.m
2279cfd782f9aterrier-plotUpdate

Last Change
Raw File

dicom_load_dictionary.m
View Options

function [tags, values] = dicom_load_dictionary(dictionary)
%DICOM_LOAD_DICTIONARY Load the DICOM data dictionary into MATLAB
% Lines in the data dictionary file have the form:
%
% # Comment
% (ABCD,EFO1) \t VR1/VR2/... \t BlanklessName \t VM \n
%
% VM can be a scalar or range (e.g., 1, 3, 1-3, 3-n)
% The data dictionary in memory is stored as a structure array with one
% entry in the struct for each attribute in the data dictionary.
% Attributes in repeating classes are expanded into the 128 attributes for
% the class.
%
% The entries of the associated 65536-by-65536 sparse lookup table are
% nonzero if a group and element combination correspond to an attribute
% in the data dictionary. The DICOM data dictionary is zero-based, so
% the indices for tag (0x0008,0x0010) are (9,17). Thus, entry (9,17) in
% the sparse array contains the index for attribute (0008,0010) in the
% struct, or 0 if it isn't in the dictionary.
% Copyright 1993-2011 The MathWorks, Inc.
if (strfind(dictionary, '.mat'))
load(dictionary);
return
end
MAX_GROUP = 65535; % 0xFFFF
MAX_ELEMENT = 65535; % 0xFFFF
values = struct([]);
count = 0;
%
% Read the dictionary.
%
fmt = '(%4c,%4c)\t%s%s%s';
[group, element, vr, name, vm] = textread(dictionary, fmt, ...
'commentstyle', 'shell');
%
% Remove duplicate entries from the list. Keep the last.
%
% Concatenate group and element to make the elements be sorted uniquely.
ge = [group element];
% UNIQUE has the property of sorting rows and keeping the last
% duplicate. This is yields the desired behavior of removing entries
% from the data dictionary which have been superceded by later entries.
[ge_sorted, idx] = unique(ge, 'rows');
% Separate the sorted group and element values.
group = ge_sorted(:, 1:4);
element = ge_sorted(:, 5:8);
% Sort and remove duplicates from the value fields using the index from
% UNIQUE.
name = name(idx);
vr = vr(idx);
vm = vm(idx);
%
% Process VR, VM
%
for p = 1:length(vr)
% VR.
if (~isempty(strfind(vr{p}, '/')))
% Multiple potential VR's.
tmp_vr = tokenize(vr{p}, '/');
for q = 1:length(tmp_vr)
if (length(tmp_vr{q}) ~= 2)
error(message('images:dicom_load_dictionary:invalidVR', vr{ p }, group( p, : ), element( p, : )))
end
end
vr{p} = tmp_vr';
else
% Single VR value.
if (length(vr{p}) ~= 2)
error(message('images:dicom_load_dictionary:invalidVR', vr{ p }, group( p, : ), element( p, : )))
end
end
%
% VM.
%
% Look for range.
h_pos = strfind(vm{p}, '-');
if (~isempty(h_pos))
% VM is a range.
if ((length(h_pos) > 1) || ...
(h_pos(1) == 1) || ...
(h_pos(end) == length(vm{p})))
error(message('images:dicom_load_dictionary:invalidVM', vm{ p }, group( p, : ), element( p, : )))
end
vm_low = vm{p}(1:(h_pos - 1));
vm_low = sscanf(vm_low, '%d');
vm_high = vm{p}((h_pos + 1):end);
if (strfind(lower(vm_high), 'n'))
vm_high = inf;
else
vm_high = sscanf(vm_high, '%d');
end
if (isempty(vm_low) || isempty(vm_high))
error(message('images:dicom_load_dictionary:invalidVM', vm{ p }, group( p, : ), element( p, : )))
end
vm{p} = [vm_low vm_high];
else
% VM is scalar.
tmp = sscanf(vm{p}, '%d');
if (isnan(tmp))
error(message('images:dicom_load_dictionary:invalidVM', vm{ p }, group( p, : ), element( p, : )))
else
vm{p} = [tmp tmp];
end
end
end % for p ...
%
% Handle the repeating classes before storing the rest.
% - All of these classes will have a group or element ending in 'xx'.
% - The spec says that new repeating classes are not to be introduced.
% Sequences will be used instead.
% - The range for these values is always 00 to ff (e.g., 5000 - 50ff).
% - Only even group values are part of the repeating group (odd ones are
% private, of course). Repeating elements, on the other hand, can be
% either odd or even.
% - Technically these classes have the same name in the data dictionary,
% but we need to uniquely identify them when placing them in the
% structure.
% - See PS 3.5-2000 Sec. 7.6 for information on repeating groups.
% - Repeating elements are an evil construction present in ACR-NEMA only.
%
% Find repeating groups and elements.
rep_group = strmatch('xx', fliplr(group));
rep_element = strmatch('xx', fliplr(element));
% Number of all attributes is (regular + repeating).
count_rep_classes = length(rep_group) + length(rep_element);
count_expanded = 128*length(rep_group) + 256*length(rep_element);
total_attr = length(vr) + count_expanded - count_rep_classes;
% Preallocate variables.
values(total_attr).Name = 'TEMP PLACEHOLDER';
values(total_attr).VR = {''};
values(total_attr).VM = 0;
rows = zeros(total_attr, 1);
cols = zeros(total_attr, 1);
% Assign values for the Repeating groups.
for p = 1:length(rep_group)
% Find the offset of the repeating group (e.g., 6000 or 5000).
offset = sscanf([group(rep_group(p), 1:2) '00'], '%x');
cur_element = sscanf(element(rep_group(p), 1:4), '%x');
% Place the repeating values in the struct.
for q = 1:128
count = count + 1;
values(count).Name = [name{rep_group(p)} '_' sprintf('%X', 2*(q-1))];
values(count).VR = vr{rep_group(p)};
values(count).VM = vm{rep_group(p)};
end
% Keep track of group and element values for the allocated attributes.
rows((count - 127):(count)) = (offset + 1):2:(offset + 256);
cols((count - 127):(count)) = cur_element + 1;
end
% Repeating elements.
for p = 1:length(rep_element)
% Find the offset of the repeating element.
offset = sscanf([element(rep_element(p), 1:2) '00'], '%x');
cur_group = sscanf(group(rep_element(p), 1:4), '%x');
% Place the repeating values in the struct.
for q = 1:256
count = count + 1;
values(count).Name = [name{rep_element(p)} '_' sprintf('%X', (q-1))];
values(count).VR = vr{rep_group(p)};
values(count).VM = vm{rep_group(p)};
end
% Keep track of group and element values for the allocated attributes.
rows((count - 255):(count)) = cur_group + 1;
cols((count - 255):(count)) = (offset + 1):(offset + 256);
end
% Don't process the repeating values twice.
group([rep_group; rep_element],:) = [];
element([rep_group; rep_element],:) = [];
name([rep_group; rep_element]) = [];
vr([rep_group; rep_element]) = [];
vm([rep_group; rep_element]) = [];
%
% Store the remainder of the values.
%
group = hex2dec(group);
element = hex2dec(element);
count = count + 1;
rows(count:end) = group + 1;
cols(count:end) = element + 1;
tags = sparse(rows, cols, 1:total_attr, MAX_GROUP + 1, MAX_ELEMENT + 1);
[values(count:total_attr).Name] = deal(name{:});
[values(count:total_attr).VR] = deal(vr{:});
[values(count:total_attr).VM] = deal(vm{:});
c4science · Help