Computer-aided bootstrap generation of characteristic curves for radiographic imaging systems

Abstract

Determination of the characteristic curve is essential for quantitative evaluation of digital as well as screen-film radiographic imaging systems. When it is not practical to generate the entire curve through variation of a single exposure parameter, bootstrap methods can be used. For the bootstrap method used here, curve segments are generated by varying one exposure parameter and multiple segments are produced at different exposure levels by varying a second parameter. The segments then are joined to form a single composite characteristic curve. If the second parameter is one for which the sensitivity of the receptor is not constant (such as x-ray tube potential or beam filtration), the shift of each segment along the log relative-exposure axis needed to join the overlapping curve sections is not known and some form of segment matching must be employed. A spline interpolation method and a polynomial fit integral method were developed for automatic segment matching and compared. For both methods, the shifts between successive segment pairs which result in optimal overlap are determined and the cumulative shifts to obtain the complete composite curve are calculated for all segments. The two methods are evaluated for several image receptors and with a known curve generated from an analytic function. Both methods closely agreed in the shifts determined for the image receptors and provided a good visual matching of the curve segments. The spline interpolation method more accurately determined the appropriate shifts for the mathematically generated curve segments. The bootstrap methods can provide complete, accurate characteristic curves, and the segment joining program makes the process fast and precise. © 2001 American Association of Physicists in Medicine.