Multivariate Regression Between Hounsfield Unit Shift, Tissue Temperature, and Tissue Contraction: A Feasibility Study of Computed Tomography Thermometry

This study aims to determine the relationship between Hounsfield Unit shift ( $\Delta {\mathrm{ HU}}$ ), tissue temperature change ( $\Delta T$ ), and tissue’s relative cross-sectional area change ( $\Delta A$ ) of ex vivo bovine liver tissues. The fresh bovine livers ( $n = 3$ ) were heated from 21 °C to 100 °C within 24 min using a custom-made ceramic hot plate and left for cooling to room temperature (~70 min). HU and tissue temperature measurement were obtained at 3-min intervals during the experiment. After that, the cross-sectional tissue area changes, and $\Delta A$ was calculated. Finally, a multivariate linear regression analysis was performed to determine the relationship between $\Delta {\mathrm{ HU}}$ , $\Delta T$ , and $\Delta A$ . The results were compared with that of the conventional $\Delta {\mathrm{ HU}}$ - $\Delta T$ linear model. Tissues that underwent a larger thermal dose experienced a higher degree of tissue contraction, an irreversible process that contributed to the non-linear behavior in the $\Delta {\mathrm{ HU}}$ - $\Delta T$ characteristic curve. This finding was in agreement with the histological test, in which the tissues nearer to the heat source had higher cell counts than those farther away. A phenomenological equation for $\Delta {\mathrm{ HU}}$ in the function of $\Delta T$ and $\Delta A$ shows better regression with the experimental data ( $R^{2} \in $ [0.838, 0.977]) than the conventional $\Delta {\mathrm{ HU}}$ - $\Delta T$ linear model. This finding indicates that tissue contraction is an important parameter to be considered in the development of CT thermometry. The $\Delta {\mathrm{ HU}}$ - $\Delta T$ - $\Delta A$ characteristic equation presented in this work offers accurate estimation of tissue temperature based on the HU change and magnitude of tissue contraction during a CT-guided thermal ablation procedure.