Frequently Asked Questions

  1. There are two guidelines for submission processes: one is at this website, and one is at STACOM 2011 website. Which one should I follow?

    2. Please only follow guidelines for the segmentation challenge provided in this website: http://www.cardiacatlas.org/web/guest/stacom2011. The information in the STACOM 2011 website was only used for call for participation. We maintain the challenge process through cardiacatlas.org website, and therefore all information in here are up to date.

  2. Why do I get the following image when I stacked binary image masks from shot-axis views in 3D?

    3.  

    The slice number values in the filename were taken from the DICOM database. In many cases, these numbers correlate with the image slice ordering in the short-axis view, but not in all cases. Therefore, unfortunately, you cannot infer the slice level ordering based on the filename format. If you stack the image slices based only on this number, then you will probably get the above pictures.

    The correct way to calculate the sequential slice level ordering from apex to base is first to get the image position information from the DICOM header. After that, you can calculate the distances between images.

  3. Why are there partial myocardium, especially at basal slices?

    4. Partial myocardium appears due to partial intersection between the model and the image slices (see images below). The ground truth uses 3D surface finite element model, which has been analyzed by experts. The 3D cardiac model has slanted surface on the base. During contraction, this slanted surface will intersect the image slices partially.

     

     

    If these partial myocardium cases are problematic to your algorithm, then you can exclude them. You will still get the same sensitivity and specificity values, because these performance parameters are measured per image slice. However, you might get slightly larger discrepancy with LV measurements, such as volume and masses, because these are measured based on the 3D model.

  4. I opened the binary images in Matlab, but why is there only a single value (= 255) from all pixels? I cannot see the myocardium in the binary images.

    5. This is actually a problem with the latest version of Matlab. If you open a PNG image using imread, you have to use 3 output arguments to get correct pixel values: 

     >> [img,~,~] = imread('DET0000101_SA1_ph10.png');

    Further information is available on the PNG section of imread documentation.

If you have still an unanswered question, please do not hesitate to contact us directly via email.