Robust correspondence of automated segmentations with manually segmented zebrafish membrane images.

Automated algorithm performance was measured against manually-segmented membrane images of the zebrafish presomitic mesoderm from four different time-points. The proposed algorithm 3 recorded an average precision of 89%, average recall of 92%, average encroachment of 0.27 m, and average volume overlap of 93%, thereby indicating an accurate performance of the method. Moreover, algorithm 3 consistently performed better than the basic watershed procedures (algorithms 1 and 2) across all the chosen metrics indicating the utility of the reconstruction procedure in improving algorithm performance.