Microscope Image Processing

that diameter x will occur, given that the cell belongs to class 1.’’ Similarly, p(xjC

2

)

is the probability that diameter x will occur, given cell class 2 (abnormal).

If we scale each of the pdfs by the a priori probability of its class, as in Figure

11.2, we get a better picture of the error situation. We could establish a decision rule

by setting a threshold value, T, on the nuclear diameter and classifying cells normal

if they fall below that and abnormal if they fall above it. The area under the dotted

curve, to the left of the threshold, is proportional to the probability of calling

an abnormal cell normal. Similarly, the area under the solid ...