How thick should CT/MR slices be to plan conformal radiotherapy? A study on the accuracy of three-dimensional volume reconstruction.

Abstract

The aim of this experimental study was to correlate the thickness of acquired CT slices (2, 4 and 8 mm) or MR slices (4 and 7 mm) with the accuracy of three-dimensional volume reconstruction as performed by a commercially available radiation therapy planning system. We used a cylindrical phantom, with a 15-cm diameter and 20-cm height, containing 5 spheres (12.7-31.8 mm diameter) of solid Plexiglas sunk in a 3% agar jelly solution. The phantom was scanned by the CT scan with 3 different slice thicknesses (2, 4 and 8 mm and a distance of 0 mm between the slices). Two different acquisition techniques (slice thickness of 4 and 7 mm with 0.8 and 1.4 mm slice distance, respectively) were compared in the MR study. The volume values calculated from measurements were compared with the known true volume values of the spheres. The average percentage volume difference between calculated and true values for the smaller spheres reconstructed with CT images 2 and 4 mm thick was generally less than 8%, whereas the error for volumes reconstructed with 8-mm-thick CT slices was more than 20%. For the large spheres, the error was generally less than 5%. The data produced by MR acquisition agreed with those obtained using CT sections. For targets less than 1.5 cm in diameter on our system it is reasonable to acquire CT images with the smallest thickness available. For targets between 1.5 and 3 cm, it seems sufficient to acquire the localization images with a slice thickness of 4 mm. For targets more than 4 cm in diameter, considering that with our radiation therapy planning system the time spent for manual contouring and for isodose calculation highly increased with the number of acquired images, we suggest that the acquisition of CT-MR slices 8-10-mm thick is totally adequate even for conformal radiotherapy treatments.