The lesion distribution for a dataset or lesions distribution specified as a one-dimensional array.
UtilLesionDistr(datasetOrLesDistr)lesDistr The lesion distribution array.
A dataset or a one-dimensional array containing the lesion distribution. For example, c(0.1, 0.2, 0, 0.7) specifies 10 percent of diseased cases have one lesion, 20 percent have two lesions, 0 percent have 3 lesions and 70 percent have four lesions. See 3rd example below.
Two characteristics of an FROC dataset, apart from the ratings,
affect the FOM: the distribution of lesions per case and the distribution
of lesion weights. This function addresses the lesions. The distribution
of weights are addressed by UtilLesionWeightsDistr. lesDistr is a
[1:nRow,2] array, where nRow is the number of unique
values of lesions per case in the dataset, 1, 2, ..., etc. The first
column of the array contains the number of lesions per case. The second
column contains the corresponding fraction of diseased cases.
See PlotRsmOperatingCharacteristics for a function that depends on
lesDistr. See TBA Chapter00Vignette2 for more details.
The underlying assumption is that lesion 1 is the same type across all
diseased cases, lesion 2 is the same type across all diseased cases,
etc. This allows assignment of weights independent of the case index.
In the third example below, relWeights = [0.2, 0.4, 0.1, 0.3] means that
on cases with one lesion the weight of lesion 1 is unity, on cases with two
lesions the weight of the first lesion to that of the second lesion is
in the ratio 0.2 : 0.4, i.e., lesion 2 is twice as important as lesion 1.
On cases with 4 lesions the weights are in the ratio 0.2 : 0.4 : 0.1 : 0.3.
There are no cases with 3 lesions in this example. Of course, on any case
the weights sum to unity.
UtilLesionDistr (dataset01) # FROC dataset
UtilLesionDistr (dataset02) # ROC dataset
UtilLesionDistr (c(0.1, 0.2, 0, 0.7)) # We specify the distribution
Run the code above in your browser using DataLab