An investigation into the problems of attenuation and area measurements made from CT images of pulmonary nodules.

Pulmonary nodules were modelled with Perspex rods of various diameters. Computed tomographic (CT) images of these rods were obtained in the centre of the scan field, both in air and in a water bath. The numerical values at the centres of the rods were compared with values obtained from scan simulations that isolated the effects of the filtered backprojection. Both real scan and simulated data were reconstructed with different algorithms. Air was found to give markedly depressed attenuation values in real scans, but not in the simulations. This effect was considered to be the result of smoothing or inappropriate beam hardening corrections. In the simulated data, air resulted in elevated attenuation values due to overshoot of the point spread function. Such overshoots varied with object size and had a more extensive influence than could be appreciated from the real scan data. Different algorithms produced different central attenuation values as a result of differing overshoot effects. These and other artifact influences make attenuation measurements taken from CT images unreliable for the assessment of pulmonary nodules. Area measurements made over a period of time may more reliably discriminate between benign and malignant nodules.