Objective: The objective of this study was to assess renal lesions measuring less than 20 Hounsfield units (HU) on noncontrast computed tomography (NCT).
Methods: Twenty-one (18.1%) of 116 consecutive pathologically proven renal cell carcinomas measured less than 20 HU on NCT and were compared with 40 confirmed benign cysts also measuring less than 20 HU. All lesions were assessed qualitatively (heterogeneous or homogenous) by 3 blinded readers and quantitatively with commercially available textural analysis software. Finally, a combined assessment was performed.
Results: Qualitative assessment performed well (sensitivity, 76%-90%; specificity, 70%-88%). Quantitative assessment revealed mean positive pixels as having the highest performance (area under the curve, 0.912; sensitivity, 90%; specificity, 80% at a cutoff value of 21). The combined assessment, using the mean positive pixel cutoff, improved the sensitivity (reader 1, 100%; reader 2, 95%; and reader 3, 95%).
Conclusion: Qualitative and quantitative assessments have relatively good performance, but the combination can nearly eliminate renal cell carcinomas being missed on NCT.