Applies the Radiological Search Model (RSM) and the Contaminated Binormal Model (CBM) ROC-curve fitting methods to 14 datasets and compares the fits to Proper ROC (PROPROC) fits obtained using Windows software downloaded from the Univ. of Iowa ROC website ca. June 2017.
ExampleCompare3ProperRocFits(startIndx = 1, endIndx = 14,
showPlot = FALSE, saveProprocLrcFile = FALSE, reAnalyze = FALSE)An integer in the range 1 to 14.
An integer in the range 1 to 14, greater than or equal
to startIndx.
If TRUE the three plots are shown along with 95
percent confidence intervals on the lowest and uppermost operating
points. The default is FALSE.
If TRUE the binned datasets are saved
for subsequent analysis using other ROC software, e.g., Windows
DBM-MRMC. The default is FALSE.
If TRUE the data is reanalyzed. The default
is FALSE in which case the previously saved results are used.
The returned value allResults is a list containing all
results from the three parametric model fits. See details.
allResults is a list-array with length equal to
(endIndx - startIndx + 1), where each element of the list-array
consists of 10 elements, see above. For example, allResults[[1]]
corresponds to
the dataset corresponding to startIndx. allResults[[2]]
corresponds to the dataset corresponding to startIndx+1, etc.
A specific member, e.g., allResults[[1]], has the following structure:
retRsm The RSM parameters following the output structure of FitRsmRoc
retCbm The CBM parameters following the output structure of FitCbmRoc
lesDistr The lesion distribution matrix
c1 The c-parameter of PROPROC
da The d_sub_a parameter of PROPROC
aucProp The PROPROC AUC
I The number of modalities
J The number of readers
K1 The number of non-diseased cases
K2 The number of diseased cases
The PROPROC parameters were obtained by running Windows software OR
DBM-MRMC 2.50 (Sept. 04, 2014, Build 4) with PROPROC and
area selected. The RSM and CBM fits are implemented
in this package. The corresponding returned objects contain all relevant
parameters. Chapter 18 of the author's book has further details.
If saveProprocLrcFile
is TRUE, the .lrc files will be written to the File-Panes
directory, overwriting any existing files with the same names.
## DPC notes on updating the results 2/17/18 ## First run PROPROC on all datasets ## 1. ret14 <- ExampleCompare3ProperRocFits(saveProprocLrcFile = TRUE) ## this generates 14 .lrc files in RJafroc ## 2. Move these files to VmWareShared folder ## 3. Start VmWare and Windows 8 ## 4. Start OR DBM MRMC, select .lrc file, select PROPROC AUC and RUN ALL ## 5. Repeat for each dataset ## 6. Move 2 files (ending with .lroc and PROPROC area pooled.csv) from ## VmWareShared to RJafroc/inst/MRMCRuns to appropriate subdirectories. ## 7. Remove spaces in names of all "proproc area pooled.csv" files ## 8. ret14 <- ExampleCompare3ProperRocFits(reAnalyze = TRUE) ## this generates new results files in RJafroc/inst/ANALYZED/RSM6 ##
Chakraborty DP (2017) Observer Performance Methods for Diagnostic Imaging - Foundations, Modeling, and Applications with R-Based Examples, CRC Press, Boca Raton, FL. https://www.crcpress.com/Observer-Performance-Methods-for-Diagnostic-Imaging-Foundations-Modeling/Chakraborty/p/book/9781482214840
Metz CE, Pan X 1999 Proper Binormal ROC Curves: Theory and Maximum-Likelihood Estimation. J Math Psychol. 43,(1):1--33.
Dorfman DD, Berbaum KS, 2000 A contaminated binormal model for ROC data: Part II. A formal model Acad Radiol 7, 427--437.
# NOT RUN {
ExampleCompare3ProperRocFits(1,1)$allResults
# }
# NOT RUN {
# }
Run the code above in your browser using DataLab