Open curve markup+cross-section analysis:an optional way of the centerline extraction and measurement of vessel?

As an exercise, I segmented manually the thrombus of a complex abdominal aortic aneurysm, and its lumen, as shown below.

It involved :

  • local threshold, for the contrasted lumen
  • threshold, for masking the thrombus intensities
  • paint the thrombus manually
  • remove small islands
  • fill holes, with the model conversion steps
  • smoothing
  • logical operation, to add the lumen and thrombus segments
  • centerline extraction

It took about 30 minutes. It should be shorter for less complex structures, or for small amount of thrombus as in your image.

My conclusion is that it’s more satisfactory than expecting a fully algorithmic method (kind of, it’s always complex code behind the scene). It may even be shorter that correcting leaks of fully algorithmic tools like Grow from seeds or Flood filling.

It’s not a magical approach, and cannot be used on a daily basis, but on a per-case basis. At least, the expected final result is there.

Regards.