Spine MRI Scans With Anomaly Labels And Reports

UniqueData's Spinal Cord Dataset contains MRI scans of the spine in DICOM format, labeled by doctors and accompanied by PDF reports. The collection includes 9 studies with images captured from different angles. The dataset was last updated on Hugging Face in September 2025.

Use Cases

• Train an anomaly detection model to identify dystrophic changes like osteochondrosis or hemangioma from DICOM image slices.
• Develop an image segmentation model to isolate pathological regions such as osteophytes or aggravated defects in spine MRI scans.
• Perform multi-task learning for joint classification of conditions like spondyloarthrosis and smoothed physiological lordosis using the provided doctor labels.
• Create a report generation or image-to-text system by linking the DICOM image features with the corresponding PDF medical reports.

Strengths

• Includes doctor-provided labels for multiple spinal conditions.
• Provides multimodal data with DICOM images paired with PDF reports.
• Offers 9 studies with images from different angles for each case.

Limitations

• Very small sample size with only 9 studies, limiting statistical power and model generalization.
• Lacks public metadata on patient demographics, scanner parameters, or exact image counts.
• Potential geographic bias indicated by a 'Regionus' tag, suggesting data may be from a single region.

Provenance

Source

UniqueData on Hugging Face.

Collection Method

MRI scans labeled by doctors, with accompanying PDF reports.

Time Range

null

Freshness

Last updated on the platform in September 2025.

Geography

Potential United States focus based on platform tag; not explicitly confirmed.