Sample datasets of mammography cases containing high resolution digitized film screen mammograms and pixel level ground truth are listed on this web page. These datasets are select cases from the Digital Database for Screening Mammography DDSM. Sets of cases with certain types of lesions have been created and listed here for you to use to determine a preliminary evaluation of the performance of your computer-aided-detection algorithm.
Volume Name | Nature of Cases | For More Information |
---|---|---|
BCRP_MASS_0 | This volume contains 39 cases that each contain at least one malignant spiculated mass. All together there are 156 mammograms in this volume of data. The purpose of this data is to serve as a training data set for algorithm development, and optimization of any and all parameters your algorithms uses. | These two datasets work together for training and testing your mass detection algorithm. These data sets concentrate on spiculated masses. Click here for more information on these datasets, how to use them and how to acquire them. |
BCRP_MASS_1 | This volume contains 40 cases that each contain at least one malignant spiculated mass. All together there are 160 mammograms in this volume of data. The purpose of this data is to serve as test data for evaluating your algorithm. Your algorithm, and all it input parameters should be fixed before using any of this data. Once you have used this data you may not use it again for estimating the performance of an algorithm. | |
BCRP_CALC_0 | This volume contains 50 cases that each contain at least one malignant lesion of clustered microcalcifications. All together there are 200 mammograms in this volume of data. The purpose of this data is to serve as a training data set for algorithm development, and optimization of any and all parameters your algorithms uses. | These two datasets work together for training and testing your microcalcification cluster detection algorithm. These data sets concentrate on clustered microcalcifications. Click here for more information on these datasets, how to use them and how to acquire them. |
BCRP_CALC_1 | This volume contains 50 cases that each contain at least one malignant lesion of clustered microcalcifications. All together there are 200 mammograms in this volume of data. The purpose of this data is to serve as test data for evaluating your algorithm. Your algorithm, and all it input parameters should be fixed before using any of this data. Once you have used this data you may not use it again for estimating the performance of an algorithm. |