Figure 6
From: Memory-efficient low-compute segmentation algorithms for bladder-monitoring smart ultrasound devices
Bland-Altman plots of the downstream volume estimation task. Since clinician annotation is considered the ground truth, it is used as the x-axis, instead of the average of the clinician annotation and the U-Net model estimate. The volume estimates in (a) (b) (c) (d) use the segmentation outputs of the FP, B+FP, and B+Q (6b, 4b) Models, respectively. The middle horizontal lines represent the average difference, i.e. bias, in volume estimation between our segmentation approach and clinician segmentation. The upper lines and lower lines indicate the 95% confidence interval of mean difference.
