BUS-UCLM: Breast ultrasound lesion segmentation dataset

The proposed dataset is comprised of breast ultrasound images from 38 patients. It consists of 683 images, of which 174 are benign, 90 are malignant, and 419 are normal. Scans were obtained with a Siemens ACUSON S2000TM Ultrasound System between 2022 and 2023. The ground truth is presented in separate files as RGB segmentation masks where green denotes benign lesions, red denotes malignant lesions, and black denotes the background or normal breast tissue. This dataset constitutes a valuable resource for research in breast cancer diagnosis, lesion detection, medical imaging, and health care applications. It facilitates researchers and practitioners to develop and examine machine learning models for the identification of benign and malignant lesions across full real cases. The segmentation annotations made by expert radiologists enable precise model training and evaluation, making this dataset a benefit in the field of computer vision and public health.

本次提出的数据集包含38名患者的乳腺超声图像，共计683张，其中良性病灶图像174张、恶性病灶图像90张、正常乳腺超声图像419张。所有扫描图像均采集于2022至2023年间，使用西门子ACUSON S2000TM超声系统完成。真值标注（ground truth）以独立文件形式提供RGB格式分割掩码（RGB segmentation masks），其中绿色代表良性病灶、红色代表恶性病灶，黑色代表背景或正常乳腺组织。本数据集为乳腺癌诊断、病灶检测、医学影像及医疗健康相关研究提供了宝贵资源，可支持研究人员与从业者开发并验证机器学习模型，用于在完整真实病例中识别良、恶性病灶。本数据集的分割标注由放射科专家完成，可支撑模型的精准训练与评估，因此在计算机视觉与公共卫生领域均具有重要应用价值。

搜集汇总

背景与挑战

背景概述

BUS-UCLM是一个专为乳腺癌研究设计的乳腺超声病灶分割数据集，包含683张标注图像（良性174/恶性90/正常419），由专业医师标注并区分病灶类型。其高精度分割掩模和真实病例数据为医学影像分析及AI模型训练提供了重要资源。