Breast Imaging Dataset with BI-RADS Annotations Organized in a Hierarchical Folder Structure

This article presents a dataset of mammographic images acquired from Oxades Company in Paraguay. The dataset is organized according to the BI-RADS (Breast Imaging Reporting and Data System) classification, a standardized system used to categorize breast imaging findings and assess the risk of malignancy. Each image is labeled according to its corresponding BI-RADS category, ranging from 0 to 6. Where the distribution of images is as follows: BI-RADS 0 (550 images) corresponds to incomplete examinations that require additional imaging evaluation. BI-RADS 1 (325 images) includes negative studies with no abnormal findings. BI-RADS 2 (1177 images) groups benign findings such as cysts or non-suspicious calcifications. BI-RADS 3 (69 images) corresponds to probably benign findings with a low risk of malignancy. BI-RADS 4 (25 images) includes cases with suspicious findings for which biopsy is recommended. BI-RADS 5 (9 images) includes studies highly suggestive of malignancy. BI-RADS 6 (8 images) corresponds to biopsy-proven malignant lesions. The dataset is structured into folders according to BI-RADS categories, facilitating its use in the development and evaluation of image processing and computer-aided diagnosis algorithms. The dataset presents class imbalance, with a higher number of benign and incomplete cases compared to malignant ones. This dataset can support researchers and clinicians in the analysis, enhancement, and automated classification of mammographic images, contributing to early detection and diagnosis of breast cancer.