Pattern Classification Methods for Analysis and Visualization of Brain Perfusion CT Maps - Computational Intelligence Paradigms in Advanced Pattern Classification - page 161

The false positive rate (the proportion of negative instances that were erroneously
reported as being positive) is:

FP

FPR = ⋅ 100 % ≈ 65 .

FP + TP

The false negative rate (the proportion of positive instances that were erroneously
reported as negative) is:

FN

FNR = ⋅ 100 % ≈ 33 .

FN + TN

Total error rate (the proportion of error instances to all instances):

FN + FP

TER = = 47 .

FP + TP + FN + TN

It can be clearly see that automatic detection of symmetry axis (ADOSM) can be
use only to simplify further manual detection of symmetry axis. The huge FPR
factor is mostly caused by overestimation of asymmetry regions (i.e. additional
regions between brain hemispheres or in the top / bottom of the perfusion map).
ADOSM can be very helpful in early estimation of visible lesions because of
computation speed and because “real” axis differs only by few degreases, but still
further manual correction may be necessary.

Table 6 Semi automatic detection results (with correction of position of symmetry axis)

Actual condition
Present Absent
Test result Positive 23 9
Negative 8 34
The false positive rate (the proportion of negative instances that were erroneously
reported as being positive) is:

FP

FPR = ⋅ 100 % ≈ 28 .

FP + TP

The false negative rate (the proportion of positive instances that were erroneously
reported as negative) is:

FN

FNR = ⋅ 100 % ≈ 19 .

FN + TN