This article presents a dataset of breast ultrasound images for the purpose of research in breast cancer detection, classification, and segmentation. The dataset consists of 780 images, each with an average size of 500 × 500 pixels, in PNG format. The images are categorized into three classes: normal, benign, and malignant. The dataset was collected from 600 female patients aged between 25 and 75 years old in 2018. The images were obtained from Baheya hospital using the LOGIQ E9 and LOGIQ E9 Agile ultrasound systems, which produce images with a resolution of 1280 × 1024. The dataset was preprocessed to remove duplicates, irrelevant information, and incorrect annotations. The preprocessing involved converting DICOM images to PNG format, cropping images to remove unnecessary boundaries, and adding image annotations to the image names. Ground truth images (mask images) were also generated using Matlab for each image to facilitate segmentation tasks. The dataset is open access and can be used for training machine learning models to classify, detect, and segment breast cancer. The dataset is the first publicly available breast ultrasound dataset, and it is comprehensive, containing breast cancer states (normal, benign, and malignant). The dataset is available for researchers interested in classification, detection, and segmentation of breast cancer. Ethical considerations were taken into account to ensure patient confidentiality and privacy. The authors also acknowledge the support of Baheya hospital and Dr. Mohamed Hamed for their assistance in managing the dataset.This article presents a dataset of breast ultrasound images for the purpose of research in breast cancer detection, classification, and segmentation. The dataset consists of 780 images, each with an average size of 500 × 500 pixels, in PNG format. The images are categorized into three classes: normal, benign, and malignant. The dataset was collected from 600 female patients aged between 25 and 75 years old in 2018. The images were obtained from Baheya hospital using the LOGIQ E9 and LOGIQ E9 Agile ultrasound systems, which produce images with a resolution of 1280 × 1024. The dataset was preprocessed to remove duplicates, irrelevant information, and incorrect annotations. The preprocessing involved converting DICOM images to PNG format, cropping images to remove unnecessary boundaries, and adding image annotations to the image names. Ground truth images (mask images) were also generated using Matlab for each image to facilitate segmentation tasks. The dataset is open access and can be used for training machine learning models to classify, detect, and segment breast cancer. The dataset is the first publicly available breast ultrasound dataset, and it is comprehensive, containing breast cancer states (normal, benign, and malignant). The dataset is available for researchers interested in classification, detection, and segmentation of breast cancer. Ethical considerations were taken into account to ensure patient confidentiality and privacy. The authors also acknowledge the support of Baheya hospital and Dr. Mohamed Hamed for their assistance in managing the dataset.