Skip to main content
. 2017 May 1;12(5):e0176636. doi: 10.1371/journal.pone.0176636

Fig 4. Optimization of tuning parameters β and γ using blocking studies.

Fig 4

Average absolute error in the estimation of VND as a function of the values for the tuning parameters β and γ for both tracers. Each point in the matrices correspond to a combination of β and γ values in the selected grids, and represents the average (across scans within the same tracer) absolute distance between the VND estimated by HYDECA, using the corresponding combination of β and γ, and VND_LASSEN. White circles indicate the optimal combinations of the tuning parameters (βopt-B, γopt-B) derived using the blocking studies, and the average absolute error in the estimation of VND in correspondence of (βopt-B, γopt-B) is reported (|errHYDb|). Green circles indicate the combinations of β and γ for which HYDECA provides an average absolute error in the estimation of VND that is higher than the absolute error committed by using the VT in the CGM as an estimation of VND (|errRR|), and the ratio between |errRR| and |errHYDb| in correspondence of the optimal β and γ is reported. The yellow and pink circle indicates the optimal combination of the tuning parameters (βopt-S, γopt-S) derived using simulation with VND_TRUE = 3 and VND_TRUE = 5, respectively, and the ratio between |errHYDs| (in correspondence of βopt-S and γopt-S) and |errHYDb| (in correspondence of βopt-B and γopt-B) is reported. VND: non-displaceable distribution volume; VND_LASSEN: VND estimated using both scans before and after blocking and Lassen plot; VT: tracer total volume of distribution; CGM: cerebellum grey matter.