BUS-BRA: 1,875 breast ultrasound images from 1,064 patients with biopsy-proven benign or malignant labels and BI-RADS (Brazil)

Provider
Wilfrido Gomez-Flores, Maria Julia Gregorio-Calas, Wagner Coelho de Albuquerque Pereira; repackaged by Chehab Lab
License
CC-BY-4.0 (source)
Size
3 files
Format
Parquet (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/chehablab/BUS-BRA)
Access
Free and public. No account with the data provider is needed.
DOI
10.5281/zenodo.8231412
Category
Health & Biomedical
Source
https://zenodo.org/records/8231412

About this dataset

BUS-BRA is a breast ultrasound (BUS) dataset of 1,875 anonymized images from 1,064 patients who underwent routine breast studies in Brazil. Every tumor is biopsy-proven: 1,268 images show benign lesions and 607 malignant ones, and each carries its BI-RADS category (2 to 5) as assigned by the radiologist. The repackaging keeps the original PNGs unchanged and adds patient ID, pathology, BI-RADS and an integer label (0 benign, 1 malignant); split by patient_id since many patients have two views. The original release also includes tumor outlines, which are not in this copy. One Parquet file under data/, about 130 MB (measured 2026-11-21); read with Hugging Face datasets. Use for benign versus malignant breast tumor classification and BI-RADS prediction from ultrasound. Original data on Zenodo (DOI 10.5281/zenodo.8231412), released under CC BY 4.0. Cite Gomez-Flores et al., Medical Physics 51(4):3110-3123 (2024).

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("chehablab/BUS-BRA")
print(ds)

Data structure

File types in sampled files: Parquet

Files

Example file: data/train-00000-of-00001.parquet

Analyze this dataset without downloading it

On Scigantic, BUS-BRA: 1,875 breast ultrasound images from 1,064 patients with biopsy-proven benign or malignant labels and BI-RADS (Brazil) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Health & Biomedical datasets