Contrast-Invariant Self-supervised Segmentation for Quantitative Placental MRI

Accurate segmentation is critical for quantitative analysis of the placenta, yet remains challenging in T2*-weighted MRI due to echo-dependent contrast variation and limited manual annotations across echoes. We propose a contrast-augmented segmentation framework that exploits the inherent diversity of multi-echo T2*-weighted MRI to learn robust, contrast-invariant representations. Our method integrates: (i) masked autoencoding (MAE) for self-supervised pretraining on unlabeled multi-echo slices; (ii) masked pseudo-labeling (MPL) for semi-supervised domain adaptation across echo times; and (iii) global-local collaboration to align patch-level features with global anatomical context. We further introduce a semantic matching loss to encourage representation consistency across echoes of the same subject. Experiments on a clinical multi-echo placental MRI dataset demonstrate that our approach generalizes effectively across echo times and outperforms supervised baselines. To our knowledge, this is the first systematic framework tailored to multi-echo placental segmentation in T2*-weighted MRI.