84 scintillations, the spatial resolution would be limit tion, obtained by taking readings everv 2 cm, shows the advantages of the narrower collimation yielding a Wie = 8.5 emvis-a-vis 11 and 15 em respectively. It 1s seen that except for the area of the distribution between 15-21 em, the calculations could be considered satis- timately by: (a) the dispersion due to the rod, (~ of its length);(b) the length of the luminesceni tron track; and (¢) the energy interval of the Cor electrons chosen for counting. The choice of the will be the result of a compromise since the resolu affected inversely by the square root of the gamn energy (that is the numberof light photonspersc1 tion), whereas the statistical information is propo. to the !» power of the numberof seintillations. factory, since the errors in f(x’) are not more than 2 few times those introduced in g(r). (Note that the tangible error at 2’ < 10 em is due to having failed to inelude g(x) readings for .« < 0, namely those bevond the spatial limits of the distribution), Before proceeding further, we must recall that even if we had perfect photomultipliers and instantaneous It is not unlikely that the choice will also dep« whether detail of the function f(x’), or accuracy integral [,° f(2’) dx’, is considered the more imp information as far as the ultimate medical pur concerned. Anillustration of what may happen 1s in Figure 66 where the information given bythe } ing curves is gathered in terms of radioactivity Integrated Activity an f(x) Interval on Fix) fem) : 0-14 14-44 44-54 eWHM = 15.0 cm h=2cem Ll FWHM = 1L0cm h=2cm LO ~99 3.2 p 27 ~10 42 4.8 . ~10 ercent Error Experimental FWHM = 85cm Activity =2% h=2cm 09 “1 3 ~7 4.6 . ~2 Value of the (uCi) <0 0.9 ~35 2.9 intervals of 14, 30, 10 em, to simulate dimensi some organs within the body. It is seen that or activity is sought over intervals larger than the the errors drop sharply within acceptable limits. A parameter of some practical importance is terval h along g(x) at which readings are taken. 4.7 ~? ure 67 the effects of intervals between 1 and 4 shown for a resolution of 8.5 em; it is seen that a: detail is concerned /. = 2 cm is best in thiscase. ] lar experiments with a resolution of 15 em (Fig: Fic. 66—Influenee of kernel resolution on accuracy of in- tegral values of f(r") over 0-14. 14-44 and 44-54 em of extended source. FWHM = 8.5 cm iN 50— ——— h=teom Jot h=2cm 1 k (\ 60— f | t | _ ; —-—n=3cm h=4em \ " | 40-— > \\ : k- z= O < 5 uy) 20 : P | r 4 ‘| = oF | | [il f ‘ Hil | ~ fy o if } aN YX # ss if ' ii i re =a Hi ‘ te fp to 0 IO 20 ft : i H , VF | J 30 | | 40 _| L 50 DISTANCE (cm) Fro. 67.—tiffect of & on details of f(x’) with Wi. = 8.5 em | 60 7