Intrinsic measurement bias on computed tomography scout view is unpredictable: Computed tomography pelvimetry using a phantom.

Our aim was to determine the degree of bias in CT scanogram measurements. We obtained standard lateral and anteroposterior (AP) pelvimetry scanograms of a phantom pelvis after placing ball bearings or aluminium rods to mark bony landmarks. Computed tomography pelvimetry was carried out at the manufacturer-recommended table height on two commercial CT scanners and at 10-mm increments up to 50 mm above and below this height. The AP inlet, AP outlet, interspinous distance and transverse diameters were each measured three times for each scanogram. The true measurements were obtained directly from the disassembled phantom. Bias was defined as the difference between the CT measurement and the true measurement. Observer error was negligible. The transverse diameter was overestimated at high table positions and underestimated at low table positions on both scanners (+6 to -10 mm). After correcting for geometric distortion, up to 6 mm bias was still present. The point at which no bias occurred was different for each scanner and did not correspond to the manufacturers' recommended table height. The outlet was overestimated on both scanners by up to 5 mm. The true inlet measurement was overestimated by 1.2 mm. The interspinous distance was minimally underestimated on both scanners. The measurements on CT scanogram were underestimated or overestimated in an inconsistent and unpredictable fashion, varying from one type of measurement to another and from CT scanner to CT scanner. This has implications for the accuracy and clinical utility of measurements obtained from a CT scanogram.