A Generalizable Framework for Kidney Stone Composition Characterization Using Dual-Energy CT.

Abstract

Rationale and objectives: Classification of non-uric acid (NUA) renal stones in dual-energy CT (DECT) is difficult due to their similar CT number ratios (CTRs) and because the CTRs change with patient size and acquisition protocol. In this work, we developed a generalizable framework to estimate correct CTR threshold for different stone types, protocols, and patient sizes and validated the results on two DECT scanners.

Materials and methods: Our framework assumes generic x-ray spectra, estimates the added filtration to match half-value-layer (HVL) measurements, and predicts the CTR of each stone type from the chemical composition and patient size. The framework was validated for four calcium or iodine inserts in two solid water phantom sizes on two DECT scanners, and on 45 human urinary stones of five types (uric acid, cystine, calcium oxalate monohydrate, brushite, and hydroxyapatite) in three different water phantom sizes on a dual-source DECT. All scans were performed at high dose, using routine acquisition parameters. The predicted CTR was compared with the measured CTR.

Results: The predicted CTRs for different stone types were consistent with experimentally measured values, with average absolute errors of 2.8% (range 1.3-4.3%), 1.8% (range 0.7-4.4%), and 1.8% (range 0.8-2.4%) for the 30, 40, and 50 cm phantom sizes. The predicted CTR errors of the four inserts were within 6.4%.

Conclusion: The developed framework uses easily obtained HVL measurements to predict renal stone CTRs of different compositions for varied patient sizes. With further refinement, it may help classify NUA subtypes in clinical scans.