meniscanData/KneeMRI-Meniscus-Dataset: MRI Meniscus Tear Dataset v1.0.0

This release contains the first public dataset for the task of automated recognition of meniscus tears in the knee joint based on MRI images.The source data were collected from patients between 2022 and 2024 at the National Scientific Center of Traumatology and Orthopedics named after Academician N.D. Batpenov (Astana, Kazakhstan). Dataset characteristics: Source file format: DICOM Number of patients: 2,242 Age: 11 to 75 years MRI scanners used: 1.5 T and 3.0 T from various manufacturers Data classes: 0 — images without meniscus tear 1 — images with confirmed meniscus tear Annotation: Manually selected and annotated: 4,262 images After expert verification: 2,000 images (1,000 without meniscus tear; 1,000 with confirmed meniscus tear) Total number of objects (annotated menisci): 7,990 Clinical features: The dataset includes only the horizontal tear type Formats and data preparation: Images exported to PNG using Weasis Only slices visualizing the menisci were selected for analysis MRI modes used: T1, T2, PD Sequences used: SE, TSE, FSE, FatSat, FS Main projections: coronal and sagittal Link to the articlehttps://doi.org/10.3390/computers14080333 How to citeWhen using this dataset, please cite: Dataset DOI: https://doi.org/10.5281/zenodo.16939444 Article DOI: https://doi.org/10.3390/computers14080333