Quantitative Evaluation of Noncontrast Computed Tomography of the Head for Assessment of Anemia.

Abstract

Anemia is an underdiagnosed clinical entity with significant mortality and morbidity. We aimed to assess whether attenuation of dural venous sinuses correlates with hemoglobin/hematocrit and to determine if the degree of anemia can be predicted by quantitative analysis of unenhanced computed tomography (CT) of the head. This is an institutional review board-approved retrospective study including 500 patients who underwent emergency department investigation for potential central nervous system etiology of their symptoms with unenhanced CT head at a tertiary care center. Computed tomographic attenuation values were obtained by 2 independent readers, whereas 2 separate investigators collected clinical data. Regression analyses were performed to evaluate the strength of correlation and the predictability of anemia and its severity on unenhanced CT. Receiver operating characteristic curve analyses were performed to evaluate sensitivity, specificity, as well as positive and negative predictive values. A total of 243 met the inclusion criteria, and attenuation values for all the dural venous sinuses were averaged and categorized according to hemoglobin values of less than 8, 8 to 10, 10 to 14, and greater than 14. Mean CT attenuation values for both readers were 36.30, 42.35, 47.99, and 53.25 Hounsfield units. Regression analysis revealed the highest positive correlation of hemoglobin/hematocrit with attenuation at the confluence of sinuses with R value of 0.63 and 0.60. The sensitivity, specificity, and negative predictive value of detecting hemoglobin of less than 10 at confluence of sinuses were 91.2%, 88.5%, and 98.6%, respectively. Interobserver agreement was found to be good (0.64) using the κ statistic. Our study substantiates direct positive correlation between CT attenuation of dural venous sinuses and hemoglobin/hematocrit, with strongest correlation at the confluence of sinuses with good sensitivity, specificity, and negative predictive value.