Minimsd: Processed 2D Medical Image Segmentation Benchmark for 10 Organs

MiniMSD provides a processed version of the Medical Segmentation Decathlon benchmark, covering 10 human organs. The dataset converts volumetric NIfTI scans into serialized 2D RGB images with corresponding segmentation masks. It is hosted by chehablab and was last updated on 2026-04-19.

Use Cases

• Benchmarking segmentation model performance based on the 10-organ medical image benchmark.
• Training 2D segmentation networks based on serialized RGB images derived from volumetric scans.
• Comparing model performance across different input resolutions based on the provided 244 and 512 pixel variants.

Strengths

• Covers 10 distinct human organs for segmentation tasks.
• Provides data in multiple resolution variants (244, 512 pixels) for flexibility.
• Derived from the established Medical Segmentation Decathlon (MSD) benchmark.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The full description is hosted externally, requiring a click-through for complete details.

Provenance

Source

Derived from the Medical Segmentation Decathlon (MSD).

Collection Method

Volumetric NIfTI scans were converted into serialized 2D RGB images with masks.

Freshness

Last updated 2026-04-19 20:04:28; freshness should be verified.