Characterization of Grey and White Matter Using Spectral Analyzis on Dual Energy CT Images

Isaac Leichter 1,2 Bezalel Fialkoff 1 Yonatan Uziel 1 Eliel Ben-David 2 Zimam Romman 3 John Moshe Gomori 2
1Physics, Jerusalem College of Technology, Israel
2Radiology, Hadassah University Hospital, Israel
3Health Care, Philips, Israel

PURPOSE: Grey matter and white matter cannot be accurately distinguished on conventional CT scans due to an overlap in their grey levels. The purpose of this study was to investigate the possibility of using Dual Energy CT (DECT) for the characterization of the two brain tissues, with the goal of generating a new image on which the two tissues are clearly distinguishable. The same methodology was explored for the early diagnosis of non-hemorrhagic stroke, as this is often difficult in conventional CT scans and therefore, MRI is required.

MATERIALS AND METHODS: DECT is able to reconstruct virtual mono-energetic (VME) images, which allows for spectral analysis, of the images obtained at different energies. Six regions of interest (ROI) in grey matter and white matter in different locations were examined in 33 cases. Regression analysis on series of VME images in the range 40-100keV revealed that the best fit for the spectral attenuation curve obtained for both tissues is a power function. Although the spectral attenuation curves of the two tissues have the same functional form, the coefficients of grey matter and white matter are significantly different (P<0.001). These coefficients were used, instead of the grey level values, to characterize the two tissues. An algorithm was developed which received as input a series of VME images. The algorithm analyzes the grey level values of each pixel in the image at different energy levels and generates a power function to fit the spectral attenuation curve of the pixel. The pixel is characterized as either grey or white matter, based on the coefficients of its attenuation function, and coloured accordingly. The algorithm then superimposes the coloured image over the original image, allowing the grey and white matter to be clearly identified.

RESULTS: The characteristic attenuation curve of both grey matter and white matter was not dependant on the location of the tissue within the brain. The algorithm was tested on three normal subjects and the resulting images accurately displayed the grey matter and white matter as expected. In four cases it was shown that following the onset of a stroke, a loss of symmetry was demonstrated, when comparing the contralateral normal brain tissue with the area affected by the stroke.

CONCLUSION: Spectral analysis of DECT images may be used to accurately differentiate between grey matter and white matter as well as providing an indication of the presence of stroke in its very early stages.

Isaac Leichter
Isaac Leichter