The goal of this study is to characterize and improve the accuracy and repeatability of 2D MRF scans in the presence of slice profile imperfections. [1]. MRF is an acquisition and processing framework that utilizes pseudorandomized acquisition scheme to enable simultaneous quantification of multiple tissue parameters, such as T1 and T2. In its original implementation, the randomized acquisition scheme were simulated into a dictionary using the nominal scan parameters such as TR and flip angle. The best fit between acquired signal and dictionary provides the underlying tissue parameters. However, because flip angle is an input variable of the dictionary simulation and affects the degree of both T1 and T2 effects on a signal timecourse, the agreement between nominal flip angle and actual flip angle is critical for obtaining accurate T1 and T2 results. Both slice profile imperfections and the local B1+ field may cause flip angles to be inaccurate. The B1+ effect varies based on location and can be corrected by pre-measuring the B1 map, for example, using the Bloch Seigert method[2].The slice profile imperfection is not location specific in 2D and thus can be simulated in the dictionary without increasing the dimensionality of the dictionary. This study simulated the slice profile of different RF pulses and validated the accuracy and repeatability of the results. The dictionary with slice profile correction required additional simulation of RF pulse and slice selection gradient. Based on the RF duration, time-bandwidth product (TBW) and slice thickness, the sinc pulse and the corresponding slice selection (SS) gradient were simulated. At the beginning of each TR, the RF pulse and slice selection gradient were first divided into segments with a 10 ms step size. The RF excitation, dephasing from the SS gradient and off-resonance, as well as relaxation were simulated continuously from one segment to the next. After all the segments were simulated, the dephasing and relaxation from the rest of the TE/TR time were simulated. Finally, the slice profile was simulated using 50 isochromats across a distance four times wider than the nominal slice thickness to account for out-slice excitation and then summed together to become the signal evolution from one pixel. A total of 287709 signal time courses, each with 3000 time points, with different sets of T1, T2 and off-resonance parameters were simulated. The total simulation time was 4.7 hours.10 cylindrical gel-phantoms with T1 and T2 ranging from 200 to 1700 ms and 20 to 110 ms, respectively, were scanned in a 3T scanner (Skyra, Siemens). Because RF duration and TBW have an impact on the MT effect [3] and slice profile, sequences with four different RF durations and TBWs were applied in separate scans to estimate the potential MT effects of the gel phantom and slice profile effect on the resulting T1 and T2 values: sinc pulse with duration of 800 us and TBW of 2, duration of 2000 us and TBW of 8, duration of 4800 us and TBW of 2, and duration of 4800 us and TBW of 8. The total acquisition time for these four sequences were 30, 40, 50 and 50 seconds, respectively. The T1 and T2 results were compared to the results from gold standard, spin-echo based measurements. The common acquisition parameters of these four sequences are: FOV of 300x300mm2, matrix size of 256x256, image resolution of 1.2x1.2 mm2, and slice thickness of 5 mm. 3000 TRs were acquired from each acquisition. For a repeatability study, both MRF scan with RF duration of 2000 us and TBW of 8 and standard spin-echo based measurements were scanned 5 consecutive days, each day with 3 repetitions. The mean and standard deviation of T1 and T2 from both methods were calculated.Figure 1 compares the T1 and T2 values from MRF, without and with slice profile simulation, to those from standard measurements. Both T1 and T2 accuracy is improved, despite the change in RF durations. Figure 2 shows the results from repeatability study, with bi-directional error bars representing the standard deviation of all measurements. MRF results are in agreement with the results from gold standard measurements, with a concordance correlation coefficient (CCC) of 0.9874 for T1 and 0.9905 for T2. In addition, both MRF and standard measurements demonstrate high repeatability. The coefficient of variance (CV) is 1.17% for T1 and 3.08% for T2 from MRF scans. This study demonstrates a slice profile correction method for 2D MRF scans. No extra scan time or processing time is needed once the new dictionary is simulated. Both T1 and T2 accuracy are improved after slice profile simulation. This work also demonstrates high repeatability of the MRF scans.