Read the Hupse-Karssemeijer LROC data file, a study comparing standalone performance of breast CAD vs. radiologists; the study actually included radiologists and residents; the following usage includes only the radiologists
DfReadLrocDataFile (RADIOLOGISTS = TRUE)Logical; if TRUE, the default, only radiologists are analyzed otherwise all readers are analyzed
The LROC dataset.
The data format is similar to the JAFROC format (see RJafroc-package)
with the crucial difference that there are two types of LL (TP) events:
those representing correct localizations and those representing incorrect
localizations. Also, every diseased case has one lesion and NLs are not possible
on diseased cases. J is one plus the number of readers. The first modality
is CAD, followed by the readers.
The return value is a list with the following elements:
NL [1, 1:J, 1:K1, 1] array containing the FP ratings
LLCl [1, 1:J, 1:K2, 1] array containing the TP correct localization ratings
LLIl [1, 1:J, 1:K2, 1] array containing the TP incorrect localization ratings
lesionNum array [1:K2], as in standard JAFROC/ROC format dataset, ones
lesionID array [1:K2], as in standard JAFROC/ROC format dataset, ones
lesionWeight array [1:K2], weights (or clinical importances) of lesions
dataType "LROC", the data type
modalityID [1:I], modality labels
readerID [1:J], reader labels
Hupse R, Samulski M, Lobbes M, et al. Standalone computer-aided detection compared to radiologists' performance for the detection of mammographic masses. Eur Radiol 2013.
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
# NOT RUN {
radData <- DfReadLrocDataFile()
str(radData)
allData <- DfReadLrocDataFile(FALSE)
str(allData)
# }
Run the code above in your browser using DataLab