There are certain patients where the 'possible dose mask' goes outside of the patient- is this intentional?
For some patients, the dose is simply in the table, so probably will not have a large effect.
However, for others (such as patient 38), there is a large ring that is far outside the body where the dose is deposited.
This would have an effect on the training algorithm and the final dose score, since the dose score calculates everything within the possible dose mask.
First off, thanks for bringing patient 38 to our attention. I'm not sure how I missed that egregious error in the final QA. We'll ensure that no such patient like that is included in the testing set that we use to determine the final rankings in the competition.
The possible dose mask is based on the “external” contour, which should be the volume of the patient that contains all the important structures (e.g., OARs, PTVs and nearby unclassified tissue). Since we used the clinical contours from several institutions there is some variability, and in a small portion of cases that variability leads to to some very odd cases like patient 38. In the majority of cases though the possible dose mask should conform to the patient's exterior. In some cases this mask will also include small pockets of air around the patient and/or the couch. We do consider these "errors", but expect them to have a very small affect on the total dose score and no effect of the DVH score.
AaronPosted by: OpenKBP @ April 15, 2020, 10:51 p.m.