LSS MRI AISSLab Dataset: Medical Spine Sagittal MRI Dataset for Segmentation and Foraminal Stenosis detection
Description
☐ Dataset: ● The LSS MRI AISSLab Dataset is a comprehensive sagittal lumbar spine MRI collection containing 500 patients with fully curated imaging data, expert clinical annotations, and detailed anatomical segmentation masks for the scientifc purpose (noncommercial). ● This dataset was approved by the Institutional Review Board (IRB) and clinically validated by neurologist experts at Firat University Hospital. The dataset includes a total of 8,500 sagittal MRI slices and 3,885 bounding-box annotations describing foraminal stenosis across the five lumbar levels (L1–L2 through L5–S1). ● A total of 1,703 right foraminal stenosis (RFS) and 1,982 left foraminal stenosis (LFS) regions were annotated. Each annotation specifies the lumbar level, anatomical side, and clinically assigned stenosis grade (Normal, Mild, Moderate, Severe). ● Across all lumbar levels, the dataset includes 1,067 Normal, 1,722 Mild, 1,049 Moderate, and 47 Severe Stenosis cases. Mild stenosis represents the largest proportion of annotations, followed by Normal and Moderate categories, while Severe stenosis is relatively rare. ● In addition to foraminal labels, the dataset provides expert-refined segmentation masks on the middle sagittal slice for key anatomical structures, including vertebrae, intervertebral discs (IVDs), sacrum, posterior A, posterior B, and the anterior background region. These masks were generated through a combined automated-and-manual refinement workflow and reviewed by neurologists to ensure anatomical accuracy. ● The dataset is organized into three components: (1) full sagittal DICOM series for each patient; (2) middle-slice images accompanied by pixel-level masks in PNG/XML format; and (3) full-slice PNG images with corresponding stenosis annotations when visible. ☐ ACKNOWLEDGMENT: This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No. RS-2023-00256517) and by the TUBITAK (The Scientific and Technological research Council of Turkey) under Grant Number: 123N325. This work was supported by the IITP(Institute of Information & Communications Technology Planning & Evaluation)-ITRC(Information Technology Research Center) grant funded by the Korea government (Ministry of Science and ICT) (IITP-2025-RS-2024-00437191). ☐ Please cite these articles: [1] Salem, Saied, Afnan Habib, Mukhlis Raza, Zaid Al-Huda, Omar Al-maqtari, Bilal Ertuğrul, Özal Yıldırım, Yeong Hyeon Gu, and Mugahed A. Al-antari. "AutoSpineAI: Lightweight Multimodal CAD Framework for Lumbar Spine MRI Assessments." In IEEE-EMBS International Conference on Biomedical and Health Informatics 2025. [2] Al-Antari, Mugahed A., Saied Salem, Mukhlis Raza, Ahmed S. Elbadawy, Ertan Bütün, Ahmet Arif Aydin, Murat Aydoğan, Bilal Ertuğrul, Muhammed Talo, and Yeong Hyeon Gu. "Evaluating AI-powered predictive solutions for MRI in lumbar spinal stenosis: a systematic review." Artificial Intelligence Review 58, no. 8 (2025): 221.
Files
Steps to reproduce
● The 3D MRI LSS dataset is well-organized and easy to work with. To reproduce and use the dataset: 1. Log in using your user credentials to access the files. 2. Download the compressed ZIP package. 3. Extract its contents, which include the DICOM data, foramina detection files, and segmentation files. 4. Load and align the 3D medical volumes with their corresponding detection or segmentation labels to begin your AI workflow.