Summary of Computation Speed for Parametric Imaging
| Number of workers | ||||||
|---|---|---|---|---|---|---|
| Parameter | Local | 2.0 | 4.0 | 8.0 | 16.0 | 31.0 |
| Time, mean (min) | 149.49 | 85.48 | 44.14 | 22.79 | 11.38 | 5.97 |
| Time, SD (min) | 0.15 | 0.12 | 0.12 | 0.08 | 0.02 | 0.01 |
| Speed-up ratio | 1.75 | 3.39 | 6.56 | 13.14 | 25.05 | |
| Efficiency (%) | 87.45 | 84.68 | 81.99 | 82.12 | 80.82 | |
Mean time required for computation was calculated from 5 repetitions of computing a parametric image for the whole volume. Local computation was used as a reference for calculating speed-up ratios by estimating parameters pixelwise on 1 core of 1 CPU without distributed computing. Efficiency was calculated as (speed-up ratio/number of workers). With regression analysis, linear relationship was found in speed-up ratio (y) vs. number of workers (x) as y = 0.812x, with r = 0.99.