Table 4 Description of some benchmarked datasets used for experimentation.

From: A generative adversarial network for synthetization of regions of interest based on digital mammograms

Database

No. of patients

No. of images

Cases of abnormalities

Description

MIAS

161

322 (MLO view of images)

All forms of abnormalities (32 shows architectural distortion)

Digitised to 50 micron pixel edge, and reduced to is 200 micron pixel edge and padded/clipped so that all the images are 1024 × 1024

Images include radiologist's truth-markings

DDSM

2620

10,480 (MLO and CC view of images)

All forms of abnormalities (approximately 137 shows architectural distortion)

The database has some associated patient information (like age at the time of study) and image information (like spatial resolution)

Images are marked with ground truth information about the locations and types of suspicious regions