The rules generated by CLIP3, while using 160 normal patient images as positive examples and 24 abnormal patient images as negative examples, each patient described by six features discretized into eight levels, are as follows:
Rule 1: F1=<2> F2=<2,3,5> F3=<2,3> F4=<2> F6=<2,3,6>
Rule 2: F2=<2,4,5,7> F4=<2,3> F6=<2,4,6>
Rule 3: F2=<3> F3=<2,4>
Rule 4: F1=<3> F4=<2,3> F5=<3>
Rule 5: F1=<3> F2=<2> F3=<3,4> F6=<3,6>
Rule 6: F1=<2> F3=<2> F4=<3>


The table below shows classification accuracy of the rules:
 
Normal Data
Abnormal data
Accuracy
Rule1
101
0
0.6793
Rule 2
75
0
0.5380
Rule3
22
0
0.2500
Rule 4
21
0
0.2446
Rule 5
8
0
0.1739
Rule 6
6
0
0.1630