Quantitative Aspects of Computed Tomography
Computed tomographic (C.T.) images are stored digitally as arrays of C.T. numbers, the elements of which are strongly related to the linear X-ray attenuation coefficients of the corresponding tissue volumes in the patient. This thesis is concerned with this quantitative information. Relating measured C.T. numbers to the properties of materials requires the ability to compute attenuation coefficients from compositions and densities. Numerical techniques have been used in the present work to produce formulae which are much simpler and more accurate than those currently used. Dual energy computed tomography permits the computation of parameters such as tissue densities and effective atomic numbers which may have more clinical relevance than the attenuation coefficient. A dual energy capability has been developed for the Wellington Hospital G.E. 8800 scanner. A simple algorithm has been developed for computing electron densities from dual energy C.T. scans. A spine mineral measurement procedure has been developed for the 8800 scanner which reports both single and dual energy bone measurements. A computationally efficient algorithm has been derived for the calculation of true mineral concentration and cancellous tissue density from dual energy C.T. measurements. Measurements on 31 volunteers compare well with published normal ranges for single energy C.T. Both single energy C.T. spine mineral densitometry and dual photon absorptiometry are prone to errors due to variable fat content, leading to ambiguity in interpreting the results of measurements on patients with anorexia nervosa. Measurements on 68 anorexic patients gave an average single energy value 18 % lower than that for the volunteers. The average mineral concentration determined from dual energy measurements was 19% lower, confirming that there is a genuine loss of mineral. Neither the single or dual energy mineral measurements showed any protective effect from the use of estrogen contraceptive pills, but both measurements correlated positively with the level of exercise at the time of examination, and with the duration of high calcium intake during adolescence. The dual energy mineral measurement showed no significant improvement in patients who had recovered from anorexia nervosa. Differences in image texture seen in C.T. images of the vertebral body reflect differences in the gross structure of the cancellous bone tissue. An in vivo study on vertebrae from 9 sheep was not useful in correlating the C.T. appearances of the cancellous tissue with scanning electron microscope (SEM) images of the same tissue because the characteristic appearances noted in humans were not present in the sheep specimens. However correlations between C.T. measurements of sheep vertebrae and trabecular bone volumes determined from SEM images of the same sections give linear regression correlation coefficients of 0.87 for the single energy measurement and 0.81 for the dual energy measurement. A study using human tissue is under way, but insufficient specimens had become available at the time of writing to be reported in detail here.