Intravoxel incoherent motion (IVIM) model has been used to study in body imaging for years^1,2,3, and this model separates true diffusion (D), fast diffusion (D*), and perfusion fraction (PF) from multiple b-value diffusion weighted image (DWI), and D* and PF reflect the microcirculatory perfusion. Due to the instability of D* and perfusion fraction, a segmented manner of processing multiple b-value was employed in the past reports^2,3 to implement it, but the stability is still needed to be improved further. Based on IVIM, this study is aimed to use an envelope bounding technique to reduce the severe signal fluctuation at low b-value area and to compare the results with segmented approach.One healthy human subject was scanned in Philips Ingenia 3.0 T MRI (Philips Healthcare, Eindhoven, The Netherlands) with dual-source RF transmit system for B1 calibration and 32 channel torso coil. Multiple b-value DWI was acquired with a navigator for respiratory compensation, and parameters were [TR=1500 ms/TE=43 ms/b-value=0,5,10,20,50,80,100,200,400,800], and it was scanned 6 times for stability analysis. The envelope bounding technique produces both upper and lower bound of the original signal, and a new signal profile is decided with [(upper bound+lower bound)/2] (Fig.1). Then this new signal profile was fitted by IVIM model, and it yields D, D* and PF, and also the original signal was fitted for comparison. Standard deviation maps were calculated from repeated scans, and three ROIs were selected for analysis (Fig.2). All the data was processed and analyzed with home-made script on MATLAB 2015a (The Mathworks, Natick, MA) platform.The signal curves shows that a less steep D* fitting and slightly lower PF with envelope bounding (Fig.1). The variation of envelope bounding is lower than the segmented one in D* standard deviation maps (Fig.2), and it is slightly lower in PF standard deviation maps (Fig.3). The improved variation decline of D* and PF of three selected ROIs is calculated in the table, and it shows an overall drop of standard deviation (Table 1& 2). The decline of D* standard deviation are 38.9%, 25.6% and 34.1% respectively in ROI 1, ROI2 and ROI 3, and the decline of PF standard deviation are 0.4%, 18.6% and 14.1% respectively in ROI 1, ROI2 and ROI 3.Results show that a reduction of variation both in D* and PF maps. The envelope bounding technique reduces the fluctuation at signals below b-value 100 (Fig.1), and it yields a more reliable fitting result of D*. Because of less steep signal profile at low b-value bounded by envelope, overall D* are lower than the segmented method. The stability of D* during repeated scans is elevated with envelope bounding due to the drop of standard deviation. There is a slight change of PF, because it is mainly derived from signals above b-value 200, and the envelope technique has a less effect in this range. The envelope bounding gives a good way to prevent the iteration of parameters jumping into wrong traps of local-minimum.The envelope bounding technique provides a stable result of D* through bounding fluctuated signals with envelopes. This technique will be tested more tasks for a precise validation, and it will be applied to hepatic disorders in the future.