Development and Testing of a Clinical Tool to Predict and Optimize Liver Contrast-Enhanced CT Imaging

Abstract

Achieving consistent and sufficient hepatic parenchyma contrast enhancement (HPCE) level can improve diagnostic performance and reduce enhancement variability; this raises the baseline image quality and optimize injection practices, both carries economic and safety implications. Patient factors, Iodine injection and scanning parameters (e.g. tube potential, scanning delay) affect HPCE in CT imaging. In this study, we developed and prospectively tested a clinical graphical user interface (GUI) tool which predicts enhancement level and suggests alternative injection/scanning parameters based on patient attributes (height, weight, sex, age). Methods: This work was based on our retrospectively-validated neural-network prediction model. We built a GUI to combine our model with an optimization algorithm, which suggests alternative injection/scanning parameters for patients with predicted-insufficient enhancement. The system was clinically-deployed and prospectively-tested in 24 patients considering a 110HU+/-10HU target portal-venous HPCE. For each patient, HPCE was calculated as the average HU-value of three ROIs and compared against the target value. Additionally, we compared the outcome with the patient’s previous similarly-protocoled scan to assess improvement and consistency. Results: The system suggested adjustment for 15 patients with median 8.8% and 9.1% reductions to volume and injection rate, respectively. All scan delays were reduced by an average of 42.6%. Comparison with previous scans shows increased consistency (CV=0.21 v. 0.11,p=0.012) while median enhancement remained relatively unchanged (111.3HU v. 108.7HU). The number of under-enhanced patients was halved, and all previously over-enhanced patients received enhancement reductions. Conclusion: We developed and tested a patient-informed clinical framework which predicts optimal patient’s HPCE; and suggests empiric injection/scanning parameters when predicted enhancement is deemed insufficient. The system improved HPCE consistency and decreased the number of under-enhanced patients as compared to their previous scans. This study demonstrated that the patient-informed clinical framework can predict an optimal patient's HPCE and suggest empiric injection/scanning parameters to achieve consistent and sufficient HPCE levels.