Digital image analysis: improving accuracy and reproducibility of radiographic measurement

Objective. To assess the accuracy and reproducibility of a digital image an alyser and the human eye, in measuring radiographic dimensions. Design. We experimentally compared radiographic measurement using either an image analyser system or the human eye with digital caliper. Background. The assessment of total hip arthroplasty wear from radiographs relies on both the accuracy of radiographic images and the accuracy of radi ographic measurement. Methods. Radiographs were taken of a slip gauge (30 +/- 0.00036 mm) and sli p gauge with a femoral stem. The projected dimensions of the radiographic i mages were calculated by trigonometry. The radiographic dimensions were the n measured by blinded observers using both techniques. Results. For a single radiograph, the human eye was accurate to 0.26 mm and reproducible to +/-0.1 mm. In comparison the digital image analyser system was accurate to 0.01 mm with a reproducibility of +/-0.08 mm. In an arthro plasty model, where the dimensions of an object were corrected for magnific ation by the known dimensions of a femoral head, the human eye was accurate to 0.19 mm, whereas the image analyser system was accurate to 0.04 mm. Conclusions. The digital image analysis system is up to 20 times more accur ate than the human eye, and in an arthroplasty model the accuracy of measur ement increases four-fold. We believe such image analysis may allow more ac curate and reproducible measurement of wear from standard follow-up radiogr aphs. (C) 1999 Published by Elsevier Science Ltd. All rights reserved.