Figure 5.

Mean accuracy assessment for the defined testing data using the multi-atlas segmentation estimate as a “silver standard”. The results demonstrate (1) the MLF framework provides a dramatic decrease in total segmentation time, (2) increasing the number of fused learners has valuable benefits in terms of segmentation accuracy, and (3) when fusing more than 5 local learners the MLF framework provides substantial and significant accuracy benefits over the joint label fusion baseline.