Multi-Disorder Annotations for Lumbar Spine Mid-Sagittal Images
Description
This dataset comprises 1038 sagittal spinal MRI images, each with a resolution of 384 × 384 pixels. The images are acquired using both T1- and T2-weighted sequences. Each image has been annotated with segmentation masks corresponding to various spinal conditions and the filenames have been standardized to ensure a clear link between raw images and their masks. Annotations have been meticulously performed by expert radiologists and spinal surgeons, ensuring high-quality, clinically relevant segmentations. The masks capture key anatomical and pathological features in each image, making the dataset valuable for developing automated diagnostic algorithms. Total number of annotations on 1038 images are 8991 (objects), on average 8.66 objects/annotations per image. The annotations are organized into distinct subfolders within the “labeled_images” directory, each corresponding to a specific spinal condition. The breakdown of objects per class is as follows:- - Normal IVD - 4131 objects - Lumbar Disc Bulge (LDB) - 2033 objects - Spinal Stenosis (SS) - 1921 objects - Degenerative Disc Disease (DDD) - 633 objects (Includes conditions disc degeneration, disc desiccation, vertebral body end-plate degeneration) - Thoracic Disc Bulge (TDB): 151 objects (Lesser number of objects since the dataset is related to Lumbar Spine Region) - Spondylolisthesis: 122 objects Each class folder contains modality-specific mask files (e.g., Sagittal_0001_T1_mask.png and Sagittal_0001_T2_mask.png), indicating that both T1- and T2-weighted images have been annotated. This dual-modality annotation enriches the dataset by providing complementary information critical for robust image analysis. Summary of Annotations (Images) is also included in class_summary.txt within the labeled_images folder.
Files
Steps to reproduce
1. Exporting mid-sagittal views both T1-weighted and T2-Weighted from the MRI dataset. (Originally taken from Sudirman, Sud; Al Kafri, Ala; natalia, friska; Meidia, Hira; Afriliana, Nunik; Al-Rashdan, Wasfi; Bashtawi, Mohammad; Al-Jumaily, Mohammed (2019), “Label Image Ground Truth Data for Lumbar Spine MRI Dataset”, Mendeley Data, V2, doi: 10.17632/zbf6b4pttk.2). 2. Loading Images in Matlab Image Labeler App (MATLAB 2024b) 3. Defining Classes including Normal IVD, Degenerative Disc Disease (DDD), Lumbar Disc Bulge (LDB), Thoracic Disc Bulge (TDB), Spondylolisthesis, Spinal Stenosis (SS). 4. Initially the annotations are performed manually by 3 volunteer trainees in Radiology with experience under 5 years. 5. In the next step, acceptance for initial annotations is given by panel of expert radiologist (2 - Radiologists) with over 15 years experience. Corrections (if required) are also endorsed. 6. Validation is performed by Spinal Surgeon with over 20 years experience.