High sensitivity in algorithm performance on synthetic data with varying noise parameters.

Algorithm performance was measured against ten synthetic datasets with progressively higher noise parameters (,) and increasing cell number. Dice refers to the area of overlap between ground-truth and automated segmentations. U, O, and M list the number of cells that were undersegmented, oversegmented, and matched respectively. Encroachment measures the average L2 Hausdorff displacement of the cell boundaries (a lower value is better). Precision measures the fraction of correctly identified cells from the automatically segmented set of cells. Recall measures the fraction of correctly identified cells from the manually segmented set of cells. The algorithm recorded a precision of at least 94%, recall of at least 98%, and an volume overlap of at least 84% even with high levels of noise, thereby indicating an accurate performance of the method.