Purpose: Renal perfusion CT can be used to assess renal lesions using continuous imaging, but can result in higher patient doses than a conventional multiphase protocol. This study investigates the effect of reducing radiation dose on quantitative imaging features extracted from the resulting image data.
Methods: Scans and raw projection data were collected from 10 patients undergoing evaluation of suspected renal cell carcinoma (RCC) via renal perfusion imaging on Siemens FORCE CT scanner as part of an IRB-approved study. Scans consist of 30 low-dose acquisitions (80kV, 100mAs) covering the kidneys beginning 8s after intravenous contrast injection. ROIs of the RCC lesion, as well as samples of the abdominal aorta and healthy kidney parenchyma were manually contoured at each timepoint. Calibrated noise was added to the projection data to simulate scans acquired at 75%, 50%, and 25% of the original radiation dose. Quantitative features (3 first-order, 3 GLCM) were calculated for each ROI across patient population and analyzed at different dose levels to assess the effect of dose reduction.
Results: The lesion ROI mean HU and GLCM-contrast feature were relatively constant across dose levels. The standard deviation and GLCM-correlation showed relatively monotonic dose-dependent behavior within the lesion ROI. Finally, the entropy and GLCM-autocorrelation features were not predictably stable. Population-wide mean of the difference between feature values at reduced dose and at the reference 100% dose increases as the dose decreases. Feature “stability” - how the features are homogeneously affected by change in dose - also decreases with lower dose.
Conclusion: This study explores the effect of reducing radiation dose in a renal perfusion protocol on several quantitative imaging features. We observed that some features are influenced by dose level, while others are not. This indicates that care must be taken to assess the impact of dose level on quantitative features.
Funding Support, Disclosures, and Conflict of Interest: This project was supported by grant funding through the Master Research Agreement between institution and Siemens Healthineers. An author is a consultant for MedQIA.