Cardiac MRI Segmentation Transfer Learning Benchmark

A 5.5 KB Excel file contains performance metrics for transfer learning experiments between two cardiac MRI datasets. Yafei Mi authored this benchmark, which was last updated in April 2026. It compares methods pre-trained on the CHD dataset and fine-tuned on the MMWHS dataset with varying labeled data amounts.

Use Cases

• Compare segmentation model performance metrics (e.g., Dice score) across different numbers of labeled patients (M) from MMWHS.
• Analyze the effect of pre-training on CHD data on subsequent fine-tuning results for cardiac structure segmentation.
• Benchmark new transfer learning strategies against reported results for tasks like left ventricle or myocardium segmentation.

Strengths

• File size is 5.5 KB, ensuring fast download and easy inspection.
• Data is shared under a permissive CC-BY-4.0 license for reuse.
• Provides a direct comparison of transfer learning performance on two established medical imaging datasets.

Limitations

• The dataset is extremely small (5.5 KB), indicating it likely contains only summary results, not raw image data or model weights.
• Specific performance metrics, column names, and the exact row count are not provided in the input.
• The underlying CHD and MMWHS datasets are not included, limiting analysis to the presented benchmark results.

Provenance

Source

figshare, authored by Yafei Mi.

Collection Method

Results from computational experiments comparing transfer learning methodologies.

Freshness

Last updated in April 2026, indicating it is a forward-dated or planned future release.