Figure 8. Longitudinal segmentation consistency comparison.

For the 29 participants with two longitudinal scans, each scan was segmented using both the multi-modality model and the single-modality model, and the volume of each subregion was calculated from the segmentation. In this figure, the volume consistency between the two scans was represented by Bland-Altman plot and intra-class correlation (ICC) for each model. The multi-modality model had smaller volume mean values and smaller standard deviations of longitudinal volume difference in all subregions. For the single-modality model, the distribution of results from poor-quality data (red dots) was more scattered. The multi-modality model provided much higher ICC values in all subregions. Higher ICC values are highlighted in bold