Use of Mechanical Turk as a MapReduce Framework for Macular OCT Segmentation

Figure 3

Representative segmentations by Mechanical Turk based manual segmentations with contrast based enhancements. Panels (a, g, m) show three raw SD-OCT macular images. Panels (b, d, h, j, n, p) demonstrate each image segmented by two different people on Mechanical Turk. Manual segmentations are shown as green lines. Panels (c, e, i, k, o, q) show local contrast based enhancement of manual segmentations as magenta lines. Panels (f, l, r) are the final consensus segmentations (green lines) after combining segmentations.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)

(k)

(l)

(m)

(n)

(o)

(p)

(q)

(r)