Hounsfield units variations: Impact on CT-density based conversion tables and their effects on dose distribution

Purpose
Determination of dose error margins in radiation therapy planning due to variations in Hounsfield Units (HU) values dependent on the use of different CT scanning protocols.
Patients and methods
Based on a series of different CT scanning protocols used in clinical practice, conversion tables for radiation dose calculations were generated and subsequently tested on a phantom. These tables were then used to recalculate the radiation therapy plans of 28 real patients after an incorrect scanning protocol had inadvertently been used for these patients.
Results
Different CT parameter settings resulted in errors of HU values of up to 2.6 % for densities of < 1.1 g/cm3, but up to 25.6 % for densities of > 1.1 g/cm3. The largest errors were associated with changes in the tube voltage. Tests on a virtual water phantom with layers of variable thickness and density revealed a sawtooth-shaped curve for the increase of dose differences from 0.3 to 0.6 % and 1.5 % at layer thicknesses of 1, 3, and 7 cm, respectively. Use of a beam hardening filter resulted in a reference dose difference of 0.6 % in response to a density change of 5 %. The recalculation of data from 28 patients who received radiation therapy to the head revealed an overdose of 1.3 ± 0.4 % to the bone and 0.7 ± 0.1 % to brain tissue. On average, therefore, one monitor unit (range 0–3 MU) per 100 MU more than the correct dose had been given.
Conclusion
Use of different CT scanning protocols leads to variations of up to 20 % in the HU values. This can result in a mean systematic dose error of 1.5 %. Specific conversion tables and automatic CT scanning protocol recognition could reduce dose errors of these types.