Figure 4

Validation of the 3D time-lapse image segmentation method. (a) MIP views of persistent dorsal abdominal muscles. (b) & (c) The performance of 3D image segmentation method for muscles 1 and 2 was determined for semi-automatic segmentation in frame k and for automated segmentation in 9 predicted frames p. Time-lapse data were acquired at 10 minute intervals. Time is shown relative to PPT. (d) Mean segmentation performance for the two muscles over 30 frames, including 3 k and 27 p-frames. Hence n_k = 6 and n_p = 54. Recall R = (TP)/(TP-FN), Precision P = (TP)/(TP-FP) and Accuracy F = 2*P*R/(P+R), with TP = true positives, FN = false negatives and FP = false positives.