Correlation Between Renal Function and Kidney Computed Tomography Histogram Analysis; A Radiomics Study

Abstract

Objective: In this study we aimed to investigate correlation between computed tomography (CT) histogram analysis of kidney parenchyma and renal functional status. Material and Method: Renal parenchyma CT densities were measured by using a region of interest (ROI) from both cortex and medulla, who had a non-contrast CT of abdomen (120 kV) and serum creatinine level in the same day in a cohort of 225 patients (115 female, 110 male). Serum creatinine levels were within normal limits in 186 and high in 39 of patients. ROI histogram data was analyzed by using Matlab® program which calculated the median, mean, maxiumum, minimum, standard deviation, variance, entropy, size % L, size % U, size % M, kurtosis, skewness and uniformity values. Results: Age and gender were significantly different among the groups (age p <0.001, gender p =0.002). Significant correlations was noted between serum creatinine level and kidney medulla mean (r=-0.22, p <0.001), maximum (r =-0,16, p <0.05), median (r =-0.17, p <0.001), size % L (r =-0.14, p <0.05) and entropy (r =-0.13, p <0.05) values as well as size %U (r =-0.14, p <0.05), size %M (r =0.14, p <0.05) density from kidney cortex. There was significant difference among patients with normal and high creatinine levels in terms of median (p =0.011), mean (p =0.004), size %M (p =0.032), and uniformity (p =0.045) values from kidney medulla as well as entropy (p =0.036) value from cortex. Conclusion: There is a negative correlation between serum creatinine levels and renal parenchyma CT densities particulary from medulla. Decreased renal function should be predicted by calculating the histogram analysis of kidney from cortex or medulla in a non-enhanced abdomen CT.