MuellerPT: Decomposition Driven Pretraining for Dense Learning in Mueller Polarimetry

2026-05-22

AI Summary

This study addresses the challenges of supervised learning in Mueller polarimetric imaging—namely, the scarcity of densely annotated data and domain shifts across samples and devices—by proposing MuellerPT, a physics-guided self-supervised pretraining method. MuellerPT introduces, for the first time, Lu-Chipman decomposition map prediction as a pretraining task to learn transferable dense representations directly from pixel-wise 4×4 Mueller matrices. Integrating polarization physics with few-shot fine-tuning, the approach achieves over a 20% improvement in DICE score for brain tissue segmentation with only 5% labeled data and an 8% gain in colorectal cancer classification accuracy with just 1% annotations. It also demonstrates strong robustness to domain shifts on human esophageal samples. Additionally, the work presents MAP-Org, the first large-scale multispectral animal polarimetric organ dataset.

Abstract

Mueller matrix imaging provides rich, physically meaningful contrast for biomedical tissue analysis, but supervised learning is hindered by scarce dense annotations and strong domain shifts across specimens and acquisition settings. We introduce MuellerPT, a physics guided pre-training approach that learns transferable dense representations by predicting Lu-Chipman decomposition maps from per-pixel 4x4 Mueller matrices. To scale pre-training, we collected a new large Multispectral Animal Polarimetric Organ dataset (MAP-Org). The pre-trained encoder is adapted with a segmentation head for grey vs. white matter segmentation in lamb brain. A classification head is used for colorectal cancer vs. non-cancer classification. Both segmentation and classification are evaluated across few-shot learning scenarios. In segmentation, MuellerPT improves label efficiency and cross specimen transfer compared to models without pre-training, achieving an absolute DICE gain of over 20% compared to the baseline trained from scratch when using 5% of the training data. In classification, MuellerPT also enhances label efficiency, improving overall accuracy by 8% compared to the baseline when using 1% of the training data. We demonstrate MuellerPT's robustness to domain shift with a qualitative evaluation of its predicted Lu-Chipman maps on an ex vivo human oesophagus sample. These results suggest that predicting Lu-Chipman decomposition is an effective and practical pretext task for robust biomedical inference from Mueller polarimetry and can pave the way for future work on label efficient Mueller imaging.

Problem

• Mueller polarimetry
• dense annotation scarcity
• domain shift
• biomedical tissue analysis
• label efficiency

Innovation

• Mueller polarimetry
• pretext task
• Lu-Chipman decomposition
• dense pretraining
• domain shift