Use this skill when the user is working with cardiac MRI data — including segmentation, reconstruction, benchmarking, or active learning for MRI acquisition. Triggers include "segment the left ventricle", "benchmark my cardiac segmentation model", "reconstruct undersampled k-space", "active sampling for MRI", "evaluate on ACDC or UK Biobank", "self-supervised learning for cardiac MRI", or "how do I report Dice for cardiac structures". Handles segmentation evaluation, reconstruction quality, dataset conventions, and active sampling strategies. Do not use for general fairness audits — use fairness-medical-imaging instead.
