Abstract
Background and purpose
In our institute, breast cancer patients undergoing adjuvant treatment are included in a protocol aimed to reduce cardiovascular morbidity (SAFE-2014, NCT2236806), assessing preclinical heart damage with heart speckle-tracking ultrasound. To develop a dose constraint related to subclinical heart damage, a reliable delineation of heart substructures based on a pre-existing guideline was made.
Patients and methods
Heart substructures of 16 left-sided breast cancer patients included in the SAFE protocol were delineated by five operators. For each substructure, a multi-contour delineation based on a majority vote algorithm (MCD) was created. A consensus-based delineation (CBD) was developed by an independent team of two blinded operators. Dice similarity coefficients (DSC) between volumes delineated by different operators and the MCD were collected and reported, as well as DSC between CBD and MCD.
Results
Mean DSCs between heart chambers delineated by each operator and the corresponding MCDs ranged between 0.78 and 0.96. Mean DSC between substructures delineated by all single operators and the corresponding MCD ranged between 0.84 and 0.94. Mean DSC between CBD and the corresponding MCD ranged from 0.89 to 0.97.
Conclusion
Results showed low inter-observer variability of heart substructure delineation. This constitutes an external validation of the contouring atlas used, allowing a reliable dosimetric assessment of these volumes within the SAFE-2014 trial.
https://ift.tt/2PSdmR7
Δεν υπάρχουν σχόλια:
Δημοσίευση σχολίου