Lumbar Spine Segmentation Dataset

This dataset comprises 1,200 sagittal-view lumbar spine images annotated with semantic segmentation masks. Images were sourced and curated from multiple open-source datasets. Of the total images, 600 were annotated from scratch and 600 were refined from previously annotated images, all finalized using Roboflow.
The dataset is provided in two versions: a raw version containing the original 1200 annotated images, and an augmented version containing 5400 images generated through data augmentation techniques. Both versions are intended to support the development and benchmarking of deep learning models for lumbar spine segmentation and related spinal pathology analysis.