Minute changes of cerebral parenchymal non-contrast computed tomography attenuation reflect cerebral volume alterations of blood.

AIMS: Non-infarcted acute cerebral ischaemic areas appear hypo-attenuated on non-contrast narrow-window computed tomography images. We aimed to determine the mechanism underlying minute computed tomography hypo-attenuation and visualise these attenuation changes on non-contrast computed tomography images.
METHODS: The cerebral parenchyma was defined by pixels with attenuation of 20-50 Hounsfield units. We calculated the mean cerebral parenchymal attenuation in non-contrast computed tomography images. We analysed the correlation of complete blood counts with corresponding mean cerebral parenchymal attenuation values using linear regression analysis. Moreover, we developed an image processing method that involved pixel colorisation based on the noise-reduced attenuation value for minute cerebral parenchymal attenuation visualisation.
RESULTS: Haemoglobin, haematocrit and red blood cell counts positively correlated with mean cerebral parenchymal attenuation values. The cerebral haematocrit is correlated with the blood haematocrit; therefore, cerebral parenchymal attenuation correlated linearly with cerebral haemoglobin concentration. Haemoglobin contents in a pixel partially determine the X-ray absorption dose and attenuation value. Pixel haemoglobin contents are determined by the cerebral volume of blood in a pixel. Image processed computed tomography images reflected cerebral volume of blood and showed the same alterations with regional cerebral blood volume maps of perfusion computed tomography.
CONCLUSIONS: Cerebral parenchymal attenuation correlated with cerebral haemoglobin concentration and cerebral volume of blood. Infarcted cerebral parenchyma shows about 5 Hounsfield units gray matter attenuation decrease. Attenuation decrease by less than 5 Hounsfield units means decreased cerebral volume of blood, indicating a reversible functional change. One cannot recognise minute hypo-attenuation (<5 Hounsfield units) in routine computed tomography images. However, it can be visualised through an image processing method on non-contrast computed tomography images. It may detect pre-infarction cerebral volume of blood and regional cerebral blood volume alterations.