Clinical polarimetry: identifying minimal Mueller subsets to increase imaging speed
File(s)
Accepted version
Author(s)
Tlemsani, Adam
Li, Yingdian
Ghosh, Abhijeet
Elson, Daniel S
Type
Conference Paper
Abstract
Polarimetric imaging offers rich contrast for tissue discrimination, yet full 4×4 Mueller acquisition remains slow and hardware intensive. We investigate which practical subsets of the Mueller matrix preserve diagnostic power, potentially enabling faster, simpler systems. Using decision tree classifiers, we benchmark reducedconfiguration models that simulate realistic hardware choices (with/without quarter-wave plates (QWPs) in the Polarisation Sate Generator (PSG)/Polarisation State Analyser (PSA); targeted row/column recovery) across five diverse biomedical domains: neurosurgical brain (PoLambRimetry), colorectal cancer (ColoPola), dermatology (Mice Skin Tumour, non-melanoma), hepatology (Liver Cancer, hepatocellular carcinoma (HCC) vs intrahepatic cholangiocarcinoma (ICC)), and breast oncology (healthy vs tumour). Across datasets, combinational subsets that omit one or both QWPs (linear PSG and/or linear PSA) often retain near-baseline performance, indicating that 3×4, 4×3 and 3x3 configurations can be highly competitive. These reductions can offer significant speed-ups (and simpler instruments/mechanics/calibration) with minimal loss in classification accuracy on several tasks. Importantly, we observe dataset-dependent variability in which subsets dominate. Practically, our findings support streamlined designs: (i) linear-only PSG/PSA systems for many classification tasks; (ii) targeted recovery of specific rows/columns prioritizing high-utility linear elements; and (iii) task-specific additions of circular states where they measurably help. This cross-domain evidence suggests that reduced Mueller polarimetry can deliver clinically relevant discrimination while cutting measurements, speeding acquisition, and shrinking system complexity, lowering barriers to adoption in clinical specimen imaging workflows.
Editor(s)
Elson, Daniel S
Gioux, Sylvain
Pogue, Brian W
Date Issued
2026-05-28
Date Acceptance
2026-04-01
Citation
SPIE Proceedings: Clinical Biophotonics IV, 2026, 14097, pp.21-21
Publisher
SPIE
Start Page
21
End Page
21
Journal / Book Title
SPIE Proceedings: Clinical Biophotonics IV
Volume
14097
Copyright Statement
2026 Published by SPIE. This is the author’s accepted manuscript made available under a CC-BY licence in accordance with Imperial’s Research Publications Open Access policy (www.imperial.ac.uk/oa-policy)
License URL
Source
SPIE Photonics Europe 2026
Publication Status
Published
Start Date
2026-04-12
Finish Date
2026-04-17
Coverage Spatial
Strasbourg, France