Figure 1.

Flowchart of the ACSAT algorithm. A, The input is the time-collapsed image $I_0$, and the output is a collection of automatically segmented ROIs. In each iteration, the Global FIBAT step identifies potential ROIs ${\left\{ROIs\right\}}_n^{\text{'}}$ by applying FIBAT, described in B, to the entire image $I_n$; and the Local FIBAT step, described in C, splits overlapping ROIs. B, Flowchart of the FIBAT algorithm. The input image is segmented using each of the test threshold values ${\tau }_1,{\tau }_2,\dots ,{\tau }_T$. The search range for the optimal threshold value (${\tau }_1,{\tau }_T$) is iteratively narrowed to contain the test threshold value which results in the maximum number of ROIs. C, Local FIBAT procedure. FIBAT, described in B, is recursively applied to each potential ROI until the resulting ROIs can no longer be separated by FIBAT.