PURPOSE: For CT dose optimization, one needs to address two important questions. The first is how various lesion-specific detection tasks demand different patient doses for the same patient. The second is how the variation of the patient size requires different patient doses for the same lesion detection task. In this study, we attempted to find quantitative solutions to these questions by utilizing a wide range of abdomen phantoms.
METHODS: A simplified model with a monochromatic fan beam passing through a bowtie-filter and an elliptical object was proposed. The model relates the minimum detectable contrast (MDC) to the size-specific dose by power index of -1/2 and to the lesion size by power index of -1 with a patient size dependence function (PSDF) as the proportionality factor. The experimental validation was performed using seven abdomen phantoms (lateral ranges: 10 cm-39 cm) scanned with helical modes at various dose levels on two 64-slice scanners (Siemens mCT and GE HD 750). Noise images were obtained using subtractions among adjacent slices in the images reconstructed with filtered backprojection. It was verified that the mean pixel value distributions from various small regions (1.8 mm-10 mm) are Gaussian, thus the concept of the statistically defined minimum detectable contrast (SD-MDC), defined as distribution's standard deviation multiplied by 3.29, can be applied. The impact of the helical pitch and the high-definition (HD) acquisition was also studied.
RESULTS: The experimental data from all phantoms were found to fit the power law well (R2 ≥ 0.983). The PSDF was found to be scanner dependent - modeled with a Gaussian amplifier (R2 = 0.983) for one manufacturer and with an exponential function for the other (R2 = 0.990). The MDC relationship was not found to be impacted by different pitches or by HD acquisition. The results were used to find the size-specific doses and corresponding acquisition techniques required by consistent low-contrast detectability for variable patient sizes. Visual comparisons on the low-contrast insert images demonstrated that the derived techniques delivered consistent low-contrast detectability.
CONCLUSIONS: We have modeled and verified the relationship of the minimum detectable contrast to the patient size, the patient dose, and the lesion size from the images reconstructed with filtered backprojection. The findings can be useful for task-specific dose modulation on abdomen CT studies.
- lesion size
- minimum detectable contrast
- patient size
- statistically defined low contrast
ASJC Scopus subject areas
- Radiology Nuclear Medicine and imaging