LUS-7Seg: A Multi-Annotation Dataset and Benchmark for Standard Plane Classification and Anatomical Segmentation in Hepatobiliary Ultrasound

LUS-7Seg is a multi-annotation dataset designed for multi-class standard plane classification and anatomical segmentation in hepatobiliary ultrasound. It contains 10,507 high-quality 2D images from 2,415 healthy subjects, covering 7 core hepatobiliary standard planes. Crucially, each image is equipped with expert-verified, pixel-level masks for the liver parenchyma and gallbladder. This dataset is intended to advance ultrasound computer-aided diagnosis from coarse image-level classification to interpretable, structure-aware localization.