Contrast Parameters for the Assessment of CT

Abstract

This study was designed a) to determine which quantitative contrast parameters for CT correlate best with lesion conspicuity obtained by a ROC analysis, b) to establish contrast and size thresholds for lesion detectability, and c) to use these contrast parameters for the assessment of previously introduced novel CT techniques, specifically subsecond/second conventional CT and spiral CT. A plexiglas GS®-phantom was designed to simulate the whole contrast range of clinical studies. A phantom with bores (0.5–3 cm) was filled with increasing concentrations of contrast medium thus exhibiting different contrast values. Dynamic CT scans of different scan times and spiral CT (Somatom Plus; Siemens AG, Erlangen) were performed with 8 mm collimations (120 kV) utilizing the same acquisition mode and reconstruction algorithm. Of the six contrast formulas the contrast-to-noise ratio (CNR) yielded the best correlation with identical scoring style as used in a receiver operated characteristic (ROC) analysis (linear correlation coefficient r = 0.881). Lesions of 0.5 cm and 1 cm diameter were always detected with a CNR > 4, in 71% with a CNR between 3 and 4, in 36% if CNR is between 2 and 3, 11% if CNR was between 1 and 2, and no lesion was detected if CNR was less than 1. Contrast values increased statistically significant (p < 0.05) with higher scan times (doses) in following order: 0.7 s < 1 s < spiral CT (1 s) < 2*1 s < 3*1 s. Spiral CT (1s) showed significant higher contrast values than a dynamic 1 s-scan but no significant higher (p = 0.115) lesion detection obtained by the ROC analysis. Only optimal acquisition times (high doses, e.g. 3*1 s) showed significant (p = 0.034) better lesion detection than spiral CT. Lesions of 0.5 cm diameter require at least a CNR of 3.42 ± 0.45 and lesions of 1 cm diameter at least a CNR of 2.91 ± 1.01 to be detectable on spiral CT. Contrast parameters allow to assess the clinical performance of novel CT techniques, measure objectively the diagnostic value of CT images, and may obviate clinical studies with the same purpose.