frocData


In this example dataset there are 100 non-diseased and 100 lesion containing mammography images. The maximum number
of lesions per diseased case is 3. Four radiologists intepreted all images using five image processing algorithms (modalities).
They marked and rated suspicious regions on an integer 1 - 5 scale, where larger ratings represented greater
confidence in presence of disease. Only modalities 4 and 5 have been retained in this dataset. The *.xlsx file can be
downloaded from <a href="http://www.devchakraborty.com/FrocData/frocData.xlsx">http://www.devchakraborty.com/FrocData/frocData.xlsx</a>. The dataset file can then be read into a dataset
object using <a rd-options="" href="/link/ReadDataFile?package=RJafroc&version=0.1.1" data-mini-rdoc="RJafroc::ReadDataFile">ReadDataFile</a> function.


A common task in medical imaging is assessing whether a new imaging system or device is an improvement over an existing one. Observer performance methodology, such as receiver operating characteristic analysis, is widely used for this purpose. Receiver operating characteristic studies are often required for regulatory approval of new devices. The purpose of this work is to software for the analysis of data acquired using the receiver operating characteristic paradigm and its location specific extensions. It is an enhanced implementation of existing Windows software (http://www.devchakraborty.com). In this paradigm the radiologist rates each image for confidence in presence of disease. The images are typically split equally between actually non-diseased and diseased. A common figure of merit is the area under the receiver operating characteristic curve, which has the physical interpretation as the probability that a diseased image is rated higher than a non-diseased one. In receiver operating characteristic studies a number of radiologists (readers) rate images in two or more treatments, and the object of the analysis is to determine the significance of the inter-treatment difference between reader-averaged figures of merit. In the free-response paradigm the reader marks the locations of suspicious regions and rates each region for confidence in presence of disease, and credit for detection is only given if a true lesion is correctly localized. In the region of interest paradigm each image is divided into a number of regions and the reader rates each region. Each paradigm requires definition of a valid figure of merit that rewards correct decisions and penalizes incorrect ones and specialized significance testing procedures are applied. The package reads data in all currently used data formats including Excel. Significance testing uses two models in widespread use, a jackknife pseudo-value based model and an analysis of variance model with correlated errors. Included are tools for (1) calculating a variety of free-response figures of merit; (2) sample size estimation for planning a future study based on pilot data; (3) viewing empirical operating characteristics in receiver operating characteristic and free-response paradigms; (4) producing formatted report files; and (5) saving a data file in appropriate format for analysis with alternate software. In addition to open-source access to the functions, the package includes a graphical interface for users already familiar with the Windows software, who simply wish to run the program.

Xuetong Zhai

RJafroc

Analysis of Data Acquired Using the Receiver Operating
Characteristic Paradigm and Its Extensions

frocData function

A dataset object consisting of a list containing 8 elements, see <a rd-options="" href="/link/RJafroc-package?package=RJafroc&version=0.1.1" data-mini-rdoc="RJafroc::RJafroc-package">RJafroc-package</a>.

Format

Zanca, F., Chakraborty, D. P., Van Ongeval, C., Jacobs, J., Claus, F., Marchal, G., & Bosmans, H. (2008).
An improved method for simulating microcalcifications in digital mammograms. Medical Physics, 35(9), 4012-8.

frocData: An example FROC dataset provided by Dr. Federica Zanca.

Description

Usage

Arguments

Format

Source