{"id":3881,"date":"2019-11-19T12:38:24","date_gmt":"2019-11-19T17:38:24","guid":{"rendered":"https:\/\/blogs.mathworks.com\/steve\/?p=3881"},"modified":"2019-11-19T12:38:24","modified_gmt":"2019-11-19T17:38:24","slug":"how-to-read-and-visualize-a-dicom-volume","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/steve\/2019\/11\/19\/how-to-read-and-visualize-a-dicom-volume\/","title":{"rendered":"How to Read and Visualize a DICOM Volume"},"content":{"rendered":"

Earlier this year, I learned something about DICOM datasets that surprised me. I had downloaded a Head-Neck CT+PET study, and I wanted to create a volume array in MATLAB. I tried to do this the hard way at first, and of course I got it wrong. (Spoiler: there's an easy way.)<\/p>

I naively assumed that there was some meaning to the order of filenames in the DICOM dataset. Here is the folder I was looking at:<\/p>

\"\" <\/p>

You can see that I had files of the form 00000.dcm, 00001.dcm, ..., 000089.dcm. So, I read these files in filename order and concatenated the data in the third dimension. And, I got a mess. I had assumed that the filename order corresponded to physical slice order, but I was wrong.<\/p>

You can see this by looking at the SliceLocation<\/tt> field returned by dicominfo<\/tt>. Let me read the files in filename order and plot the resulting slice locations.<\/p>

folder = \"\/Users\/eddins\/OneDrive - MathWorks\/General Reference\/D\/DICOM Head-Neck-PET-CT Dataset\/HN-CHUM-001\/08-27-1885-PANC. avec C.A. SPHRE ORL   tte et cou  -TP-74220\/3-StandardFull-07232\"<\/span>;\r\nd = dir(folder + \"\/*.dcm\"<\/span>);\r\nfor<\/span> k = 1:length(d)\r\n    info = dicominfo(folder + \"\/\"<\/span> + d(k).name);\r\n    slice_location(k) = info.SliceLocation;\r\nend<\/span>\r\nplot(slice_location)\r\ntitle('Slice location'<\/span>)\r\n<\/pre>\"\" 
Uh oh. That's not good. Does that even make sense, I wondered? Are these all unique slice locations? Let me try sorting them, first, and then plotting them.<\/p>
plot(sort(slice_location))\r\ntitle('Sorted slice locations'<\/span>)\r\n<\/pre>\"\" 
That's more like what I expected. But it means that my assumption about the filenames was completely off.<\/p>
But wait! As I said, I was doing this the hard way. The easy way is to use the function dicomreadVolume<\/tt>, which uses the DICOM metadata to automatically figures out how to arrange the slices in the output volume.<\/p>
[V,sp] = dicomreadVolume(folder);\r\n<\/pre>
whos V<\/span>\r\n<\/pre>
  Name        Size                     Bytes  Class    Attributes\r\n\r\n  V         512x512x1x90            47185920  int16              \r\n\r\n<\/pre>
The size of the third dimension is 1 (we call that a singleton<\/i> dimension) because the voxels do not have multiple color components. We can use squeeze<\/tt> to get rid of it.<\/p>
V = squeeze(V);\r\n<\/pre>
whos V<\/span>\r\n<\/pre>
  Name        Size                   Bytes  Class    Attributes\r\n\r\n  V         512x512x90            47185920  int16              \r\n\r\n<\/pre>
The second output of dicomreadVolume<\/tt> contains spatial information about the voxels.<\/p>
sp\r\n<\/pre>
\r\nsp = \r\n\r\n  struct with fields:\r\n\r\n       PatientPositions: [90×3 double]\r\n          PixelSpacings: [90×2 double]\r\n    PatientOrientations: [2×3×90 double]\r\n\r\n<\/pre>
I like to use Volume Viewer<\/b> (volumeViewer<\/tt>) to figure out the best ways to visualize a volume.<\/p>
volumeViewer(V)<\/pre>
It is helpful to use the spatial information returned by dicomreadVolume<\/tt> for Volume Viewer<\/b>.<\/p>
slice_location = sort(slice_location);\r\nscale_factors = [sp.PixelSpacings(1,:) slice_location(2)-slice_location(1)];\r\nvolumeViewer(V,'ScaleFactors',scale_factors)<\/pre>
Here's a Volume Viewer<\/b> screen shot with this data:<\/p>
\"\" <\/p>
From the Volume Viewer<\/b>, you can save rendering and camera configuration information to a struct that you can pass to volshow<\/tt>, which displays the volume visualization in ordinary figure window. You still have to pass in the scale factors as a separate argument.<\/p>
volshow(V,config,'ScaleFactors'<\/span>,scale_factors);\r\n<\/pre>\"\" 
Do you work with DICOM volume data in MATLAB? Let us know in the comments what you would like to see.<\/p>
Data credits<\/b><\/p>
  • Cancer Imaging Archive: Clark K, Vendt B, Smith K, et al. The Cancer Imaging Archive (TCIA): Maintaining and Operating a Public Information Repository. Journal of Digital Imaging. 2013; 26(6): 1045-1057. doi: 10.1007\/s10278-013-9622-7.<\/li>
  • Head-Neck-PET-CT collection: Vallières, M. et al. Radiomics strategies for risk assessment of tumour failure in head-and-neck cancer. Sci Rep 7, 10117 (2017). doi: 10.1038\/s41598-017-10371-5<\/li>
  • Subject ID: HN-CHUM-001<\/li><\/ul><\/div>