Impact of computed tomography noise and region of interest sizes on the characterization of homogeneous renal mass: a proof-of-concept study.

BACKGROUND/PURPOSE: Computed tomography (CT) attenuation plays a central role in characterizing homogeneous renal masses, but image noise may introduce variability in CT number measurements. This study investigated whether CT noise contributes to mischaracterization of homogeneous renal masses and examined the influence of region-of-interest (ROI) size on this phenomenon. METHODS: CT images of a simple cyst, a hemorrhagic cyst, and a papillary renal cell carcinoma (RCC) from 2 patients were analyzed. Circular ROIs of five sizes (1 pixel, .1 cm2, .3 cm2, 1.0 cm2, 3.0 cm2) were used to measure CT numbers on pre- and postcontrast images. Using a 10-Hounsfield unit (HU) enhancement threshold, instances of mischaracterized ROIs were recorded and the corresponding experimental mischaracterization probabilities were calculated. Estimated mischaracterization probabilities were derived from enhancement threshold, measured enhancement, and image noise. Linear regression was used to evaluate the association between experimental and estimated mischaracterization probabilities. RESULTS: Experimental mischaracterization probability decreased with increasing ROI size, reaching .000 for simple and hemorrhagic cysts at ROI ≥1.0 cm2 and for papillary RCC at ROI ≥ .1 cm2. Linear regression showed a significant positive correlation between the experimental and estimated mischaracterization probabilities (R2 = .946, regression coefficient = .763, p < .001). CONCLUSION: CT noise may contribute to threshold-based mischaracterization of homogeneous renal masses, particularly with smaller ROIs.