Pediatric Multiple Sclerosis OCT Scans with 52 Segmented Features

211 optical coherence tomography scans from pediatric multiple sclerosis patients and 52 scans from children with non-inflammatory conditions. The dataset includes raw 3D macular and optic nerve head images plus 52 automatically segmented features, collected prospectively through the Neuroinflammatory Registry of the Hospital for Sick Children. Chaojun Chen published the data in 2026 to support AI-driven diagnostic tool development.

Use Cases

• Training multimodal deep learning models for pediatric MS diagnosis based on raw OCT images and segmented features.
• Comparing early and late fusion model architectures for medical image classification using the described multimodal approach.
• Benchmarking classical ML classifiers like SVC against deep learning models (e.g., ResNet, DenseNet) on OCT-derived features.
• Studying the structural integrity of the retinal nerve fiber and ganglion cell layers in pediatric neuroinflammatory conditions.

Strengths

• Includes 211 scans from pediatric multiple sclerosis patients and 52 scans from a control group of 29 children with non-inflammatory conditions.
• Provides both raw 3D OCT images and 52 automatically segmented structural features for multimodal analysis.
• Dataset is associated with a study where an early fusion model achieved an AUC of 0.90 and 87% accuracy.

Limitations

• Dataset size is 401.1 KB, indicating a very limited scope of data, likely containing summary or supplementary files rather than the full image set.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale model training.

Provenance

Source

Neuroinflammatory Registry of the Hospital for Sick Children (REB#1000005356).

Collection Method

3D OCT scans collected prospectively through a clinical registry.

Freshness

Last updated 2026-04-21 04:11:04; freshness should be verified.