Source of Support: None, Conflict of Interest: None
Curve filtration in nuclear medicine is a common technique for quantitation of certain physiological parameters. The most common and acceptable methods are 3 point and 5 point smoothing, in which the value of each element is replaced with a symmetrically weighted average of the element itself and some juxtaposed points. Although this procedure is simple, selection of the number of filter elements and correct weighting factors is an extremely difficult task. Both the real signals and noise have a wide range of frequencies, which generally have some overlapping components. Removal of noise inevitably removes the common frequency components of the data and results in distortion of useful data. Incorrect weighting factor either leaves some noise untouched or unnecessarily distorts the data. We have devised an algorithm (window filter), which defines two nested windows (narrow and wide) around each element of the data. The location of each element is compared with the window limits before replacing its value by 3point smoothing. The points within narrow window are left untouched and those outside it are subjected to 3 point smoothing.