Accuracy of weighted CTDI in estimating average dose delivered to CTDI phantoms: An experimental study

The concept of the weighted computed tomography dose index ( $$ ) was proposed in 1995 to represent the average CTDI across an axial section of a cylindrical phantom. The purpose of this work was to experimentally re-examine the validity of the underlying assumptions behind $$ for modern MDCT systems.

To enable experimental mapping of $$ in the axial plane, in-house 16 and 32 cm cylindrical phantoms were fabricated to allow the pencil chamber to reach any arbitrary axial location within the phantoms. The phantoms were scanned on a clinical MDCT with five beam collimation widths, three bowtie filters, and four kV levels. To evaluate the linearity and rotational invariance assumptions implicitly made when the weighting factors of 1/3 and 2/3 in the $$ formula were originally derived, $$ was measured at different radial and angular locations within the phantom for different collimation, bowtie, and kV combinations. The average CTDI ( $$ ) across the axial plane was calculated from the experimental two-dimensional (2D) dose distribution and was compared with the traditional $$ .

For both phantoms under all scan conditions, the axial dose distributions were found to have significant angular dependence, potentially due to the x-ray attenuation by the patient couch or the head holder. The radial dose profiles were also found to significantly deviate from linearity in many cases due to the presence of the bowtie filter. When only the 12 o'clock peripheral $$ and the traditional weighting factors were used to calculate $$ , the average dose was overestimated in the 16 cm phantom by up to 8.4% at isocenter and up to 35.3% when the phantom was off-centered by 6 cm; in the 32 cm phantom at isocenter, the average dose was overestimated by up to 12.8%. Using an average of the four peripheral $$ measurements at the 12, 3, 6, and 9 o'clock locations reduced the error of $$ to within 1.2% in the 16 cm phantom. For the 32 cm phantom, even by using the average of the peripheral measurements, the traditional $$ underestimated the average dose by up to 4.3% due to aggressive drop-off of the $$ at the phantom periphery.

The linearity and rotational-invariance assumptions behind the traditional $$ formalism may not be valid for modern CT systems and thus $$ may not accurately represent the average dose or radiation output within a CTDI phantom. Utilizing data from all four peripheral locations always improves accuracy of $$ in representing the true average dose. For the large (32 cm) phantom, nonlinear models and more measurement points are needed if a more precise estimation of the average axial dose is required.