When researchers are designing and testing software they need to test their software against a 'golden standard'. The 'golden standard' is a standard which has proven to be accurate.
For example, when testing the automated CTC they test it against known results of a CTC that a radiologist has studied.
Correct with respect to (wrt) golden standard = True Positive (TP)
(System found Polyp and it agreed with Radiologist)
Missed wrt golden standard = False Negative (FN)
(System didn't detect polyp but radiologist did)
Found and does not agree with golden standard = False Positive (FP)
(System detected polyp but no polyp found by radiologist.)
 | Sensitivity of the design = Probability of finding a positive result given that the condition under consideration is true. |
-> (# TP /# TP + # FN)
| False Negative Rate = Compliment of sensitivity. Probability the feature is said not to be there given that it is. |
-> (1 - PTP)
|
False Positive Rate = Probability the feature is said to be there given that it is not. |
|
Specificity (True Negative Rate) = probability of a negative result, given that the condition under consideration is said to be false. Compliment of the Probability of False Positive. |
-> # TN / ( # FP + # TN )
|
Predicitve Value Positive ( PVP) = Probability feature is there given that it is said to be there. |
-> # correctly marked / Total # marked
-> # TP / (# TP + # FP )
Probability is important to researchers when they are testing their software as they can calculate how accurate their system is.