Table 5.
Considerations when choosing a method.
| Method | Spatial correspondence | Volumetric correspondence | Lesion load | Different field strength | Different scanners | Computational Time |
|---|---|---|---|---|---|---|
| Cascade | − | − | − | − | +/− | ++ |
| kNN-TTP | + | ++ | + | + | + | + |
| Lesion TOADS | − | +/− | − | + | − | +/− |
| LST-LGA | − | +/− | − | + | + | +/− |
| LST-LPA | +/− | ++ | + | +/− | +/− | + |
Note: ++: highly recommended; +: recommended; +/−; neutral; −: not recommended. Spatial correspondence: based on Dice’s Similarity Coefficient (DSC). Volumetric correspondence: based on intraclass correlation coefficient (ICC) and mean and mean absolute WMH volume differences. Lesion load: based on both spatial and volumetric correspondence with varying lesion loads. Different field strength: based on both spatial and volumetric correspondence on 1.5 Tesla compared to 3 Tesla MRI scanner of the same MRI vendor. Different scanners: based on the variation in performance across scanners, both in terms of spatial and volumetric correspondence. The (qualitative) recommendations were based on the results of the present study.