Table 2. Minutes per year with the corresponding probability and the inverse cumulative probability as a function of standard deviation. In the third step, the value rqh, which represents the same range for an allowed normal distribution of the quarter-hourly average frequency deviation, is calculated based onthe assumption that the two signals are not correlated: In the fourth step, the ranges which correspond to the probabilities required bySO GL Article 128(3) are calculated taking rqh as basis. The probabilities are calculated as follows: For the calculation of the ranges, the inverse cumulative probabilities of and will be used. SO GL Parameters qh per year Probability inverse cumulative probability value as c σ qh outside level 1 ACE range 5256 0.85 1.0364 σ qh outside level 2 ACE range 876 0.975 1.96 σ Table 3: Values outside the ranges. In the last step, the level 1 and level 2 ranges (L1 and L2) are calculated for each LFC BLOCK. With KSA as K-Factor of the Synchronous Area expressed in MW/Hz, KFCR as the total FCR of the Synchronous Area and KFCR,i as initial FCR obligation of LFC BLOCK i, the targets are given by:

Table 2. The fact that MTT values in our study were less affected than were CBF and CBV values when the dose was low- ered is in concordance with the results of a study by Xxxxxxx et al (21) in which several types of reconstruction algo- rithms were compared when the total CBF 0.911 ,.0001 .094 .118 dose was decreased by one-half. Those CBV 0.906 ,.0001 .025 .089 authors found, however, that the CBV MTT 0.864 ,.0001 .075 .371 values of gray matter actually were un- Gray matter derestimated by 18.6% in all patients, CBF 0.895 ,.0001 .408 .923 and the CBV values of white matter CBV 0.917 ,.0001 .094 .477 were overestimated by 1.5% with fil- the CT perfusion maps, authors of two studies (18,19) reported that clinical decision making is unlikely to change if the overall variability stays within 10%. Dose reduction to less than 2.5 mSv results in overestimation of patient perfusion values, particularly of CBF and CBV values. We hypothesize that increased noise levels lead to increased small-scale gradients on the intensity curves, which are erroneously inter- preted as increased blood flow and vol- ume during the deconvolution process inherent in the perfusion analysis. Ac- cording to the central volume principle (20), MTT is less affected. tered back projection as the recon- struction algorithm, but they gave no explanation for this observation. Our optimal point of dose reduction was lower than the reduction of 33% (from 190 mAs to 125 mAs, no effective dose values were reported) suggested by Ju- luru et al (22). These authors found lit- tle effect of lower dose settings on per- fusion values, although in their study, Xxxxxx et al investigated five dose set- tings with simulated noise, whereas we used real measured noise. To our knowledge, previously pub- lished work is limited primarily because of the difficulty of realistic simulation of patient data at different dose levels. In other studies, this was achieved by simply adding Gaussian noise (22), by reconstructing only a part of the total number of raw projections (21), or by using dual-source CT scanners and let- ting the two tubes operate at different milliampere and kilovolt settings simul- taneously during acquisition (23). An alternative to our approach would be to use dedicated low-dose simulators (24,25). Our approach combined real patient tissue curves with multiple CT scans at a wide range of tube current settings to construct patient-specific digital phantoms. In van den Boom et al (14) these patient-spe...