Table 2-. Results of focused classification on an asymmetric unit when the mask is applied on the wrong region.

Classification was performed after application of a 2-D mask or 3-D mask onto a leg (see Figure 2B), while the heterogeneity was characterized by the mobility in the arms/hands (scenario 1), and the κ coefficient was evaluated for the five SNRs and for each mask. Whereas the 2-D masking displayed some “leakiness” at the highest SNRs, the 3-D masking showed completely random classification.

SNR	2-D mask	3-D mask
0.100	0.23	−0.01
0.050	0.11	0.00
0.025	0.01	0.01
0.013	0.00	0.01
0.006	0.00	0.00
pure noise	−0.01	0.00