Comparison of the information content of data-blended and conventional photoscans.

Abstract

In 1965 data-blended rectilinear scans were suggested (4) in an effort to improve the appearance and fidelity of conventional photoscanning. This method had the obvious advantage of eliminating the conventional “raster” effect and smoothing spurious background events to a level at which they were no longer apparent. While these improvements in the appearance of the scan favor the data-blended method, the question of relative information content of each system has not been established. Considerable subjective comparison failed to answer this question, so an objective mathematical comparison was undertaken to resolve the inconclusive empirical studies illustrated in Figure 1. Since radiation is a wave form, any complex radiation distribution can be represented by the summation of pure sine waves of various frequencies, in the same way that the complex sound of an orchestra can be reduced to a summation of sine waves. Thus, a sinusoidal radiation distribution was selected and then mathematically scanned by models simulating the conventional and the data-blended methods. Figure 2 shows the assumed radiation distribution. The conventional scan model is illustrated in Figure 3, which demonstrates how deviations from the actual distribution D(x) result from the averaging effect of the collimator. Point B, for example, which has an actual value of zero, is represented by B′ with a value greater than zero because the scanning collimator sees not only point B but points immediately adjacent to point B which have values slightly greater than zero. Their contributions result in the average value B′. A similar evaluation of all points along the actual sine wave D(x) results in the conventional scan distribution Q(x). It is obvious that the degree of deviation is determined by the width seen by the collimator: a narrow or fine focus collimator would thus result in a smaller degree of deviation (Fig. 4). There are practical limitations, however, to the collimator width since the amount of time needed to make a scan must increase as the collimator width decreases. Since it is unrealistic to expect a patient to maintain the same position for more than thirty to forty-five minutes, very fine focus collimators cannot be used. Data-bended photoscans can be similarly expressed by mathematical models. When the original radiation distribution D(x) is scanned by this method, the resulting blended distribution is represented by Qb(x) (Fig. 5). (Details of this mathematical model are available in the supporting Appendix.) We are now in a position to compare the results of these two methods by measuring how much deviation from the original distribution exists. For both methods of scanning, a family of sinusoidal distributions was calculated for collimators with values in the clinically useful range.