[Quantitative evaluation of blood in vivo with computed tomography]

A positive correlation between the attenuation values of blood in vitro and hemoglobin concentration has been described by other authors. Our study is aimed at: 1) investigating the influence of artifacts on the measurement of blood density in vivo, 2) verifying the possible correlation between blood values and Hounsfield's units (HU) in vivo, 3) assessing whether such a correlation may be useful in calculating hemoglobin concentration, 4) investigating whether a significant difference in blood density exists between normal and anemic subjects, and 5) investigating the eventual borderline density values which separate the two populations. Fifty adult patients underwent CT of the upper abdomen for several clinical reasons. Twenty-six males, mean age 63 +/- 15 SD, and 24 females, mean age 63 +/- 14 SD, were included in our series. The two groups were then subdivided into two additional groups including anemic and normal subjects -14 g/dl and 12 g/dl being taken as the borderline values for males and females, respectively. Blood density was measured after defining two regions of interest (ROI) on the aorta and inferior vena cava. The results show that: 1) there are few image artifacts, which are sometimes detected only because of different numbers in aorta and vena cava; for quantitative blood measurement, 4 HU was considered as an acceptable difference; 2) the correlation between HU and hemoglobin concentration which was found in vitro was also seen in vivo (r = 0.76, p < 0.001). Furthermore, high positive correlation was observed between attenuation values and red cell count (r = 0.68) or hematocrit (r = 0.75). 3) However, this correlation cannot be used to calculate the patients's blood count since HU dispersion relative to the regression line is too high. The straight line of regressed computation, reporting the density values on the abscissa axis, shows a b = 0.33 slope and the intersection point is a = -1.43, SE b = 0.04, SE a = 1.76 and SE estimate = 1.56. When the mean density value was 43 HU in our sample, the expected hemoglobin value was 12.76 g/dl (+/- 1.58 SD). 4) Nonetheless, a significant difference in blood attenuation values was seen between normal subjects and anemic patients (p < 0.001). 5) This difference identifies a borderline density value, below which the patient is sure to be anemic (99% confidence), but nothing else can be said on the patient's normality: the value is 33 HU for females and 36 HU for males. These values allowed 20% of anemic subjects to be correctly identified, versus 5% of the same group detected with diagnosis images.