Variability in target volume delineation on CT scans of the breast

To determine the intra- and interobserver variation in delineation of the target volume of breast tumors on computed tomography (CT) scans in order to perform conformal radiotherapy.

The clinical target volume (CTV) of the breast was delineated in CT slices by four radiation oncologists on our clinically used delineation system. The palpable glandular breast tissue was marked with a lead wire on 6 patients before CT scanning, whereas 4 patients were scanned without a lead wire. The CTV was drawn by each observer on three separate occasions. Planning target volumes (PTVs) were constructed by expanding the CTV by 7 mm in each direction, except toward the skin. The deviation in the PTV extent from the average extent was quantified in each orthogonal direction for each patient to find a possible directional dependence in the observer variations. In addition, the standard deviation of the intra- and interobserver variation in the PTV volume was quantified. For each patient, the common volumes delineated by all observers and the smallest volume encompassing all PTVs were also calculated.

The patient-averaged deviations in PTV extent were larger in the posterior (42 mm), cranial (28 mm), and medial (24 mm) directions than in the anterior (6 mm), caudal (15 mm), and lateral (8 mm) directions. The mean intraobserver variation in volume percentage (5.5%, 1 SD) was much smaller than the interobserver variation (17.5%, 1 SD). The average ratio between the common and encompassing volume for the four observers separately was 0.82, 0.74, 0.82, and 0.80. A much lower combined average ratio of 0.43 was found because of the large interobserver variations. For the observer who placed the lead wire, the intraobserver variation in volume was decreased by a factor of 4 on scans made with a lead wire in comparison to scans made without a lead wire. For the other observers, no improvement was seen. Based on these results, an improved delineation protocol was designed.

Intra- and especially interobserver variation in the delineation of breast target volume on CT scans can be rather large. A detailed delineation protocol making use of CT scans with lead wires placed on the skin around the palpable breast by the delineating observer reduces the intraobserver variation. To reduce the interobserver variation, better imaging techniques and pathology studies relating glandular breast tissue to imaging may be needed to provide more information on the extent of the clinical target volume.