The Effect of Image Magnification and Context Reduction on Diagnostic Performance with a Digital Workstation

Summary

In this paper the influence of image presentation on digital workstations for radiological images was studied concerning magnification of details in chest images with and without contextual information. Five experienced radiologists had to report digitized chest radiographs with the diagnosises “no disease”, “lung nodules” and “thin opacifications”. Image reviewing was done under five different magnification conditions: (1) no magnifications allowed, (2) one fixed sized region of interest (ROI) without context information, (3) two ROI simultaneously, (4) ROI with a reference image, (5) two regions of interest simultaneously with a reference image.

For the evaluation of the different viewing conditions we measured the diagnostic performance and the time for different steps in the reporting session. Furthermore subjective comments were collected. Quantitative parameters of the diagnostic performance were analysed by ROC and Cohen’s Kappa statistics.

There was no statistical significance between condition (1) and conditions (2) and (3). The diagnostic performance, however, is significantly improved by magnification with a reference image at the same time (4) and (5).

For the improvement of the diagnostic performance by magnification of a region of interest, the reference image is significant. It is an essential result of our investigation that a reference to the complete image will markedly improve the diagnostic performance in the case of any region of interest operation (like the above mentioned magnification).