Quantitative CT tomography for radiotherapy treatment planning: calibration phantom and sources of error

Since tissue inhomogeneity correction in radiotherapy treatment planning systems is based on electron densities determined by CT scanning, a correct scanner calibration is mandatory to achieve a 2% accuracy in electron density determination. A quality assurance programme on quantitative CT scanning must include: CT number to electron density conversion calibration for both soft tissues and bone; check of calibration constancy in time and spatial uniformity; dependence on kVp, mAs and slice thickness; dependence on patient size, shape and position in the field of view (FOV); geometrical accuracy; choice of appropriate tissue substitutes. For this study we made a cylindrical PMMA phantom (28 cm in diameter) with 21 holes for insertion of inhomogeneities. We used organic liquids and water as soft tissue substitutes and K2HPO4 aqueous solutions for bone. Organic substances containing high Z elements (Cl or S) resulted unsuitable as their effective atomic number (for photoelectric absorption) was too high compared to that of bone (ICRU n.44). The standard calibration was made with the circular phantom placed in the centre of the FOV. Then slices were taken displacing the phantom off-centre, within the FOV or partially outside it. Two ears were added to the sides of the phantom to change the shape (axis ratio 1.5) and size. While the CT number/electron density conversion curve for soft tissues is not influenced by scanner related parameters, that of bone is strongly related to kVp and scanner model. CT numbers were affected by the change in size and shape and phantom displacement partially outside the FOV. These variations were greater near the added ear or the crossing between phantom and FOV. All significant CT number variations (absolute value greater than 20) were negative (reduced density) and strongly related to electron density: less than 40 for soft tissue substitutes, up to 112 for bone substitutes