# Breast Cancer US Classification | ML4IM 2025-2026 — Data

来源：https://www.kaggle.com/competitions/breast-cancer-us-classification-ml-4-im-2025-2026/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

The training set contains two folders, corresponding to the "benign" and "malignant" classes of interest. This folder separation provides information regarding the label of each study (which is also the target of our predictions).

Each breast ultrasound study is referred to with a numerical *Id* (e.g. "benign (25)") and has a corresponding lesion segmentation (e.g. "benign (25)_mask"). Each study has at least one corresponding segmentation file, containing the mask of a lesion.
However, there are studies with more than one segmented lesion, and each segmentation mask is reported in a different file (e.g. "benign (25)_mask" and "benign (25)_mask_1"). In these cases, all the lesions belonging to the same study share the same class (e.g. both lesions of study "benign (25)" are benign). 
Studies and masks are provided in *png* format.