A Pipeline for Chamber-Resolved Analysis of Pore Traits in Foraminiferal µCT Volumes
Wu, Hanqing LU
; Choquel, Constance
; Ni, Sha
; Filipsson, Helena L.
LU
and Pirzamanbin, Behnaz
LU
(2026)
In Lecture Notes in Computer Science
1700.
- Abstract
- Foraminiferal shells are abundant in marine sediments, well preserved, and morphologically sensitive to environmental conditions, making them widely used archives of past marine environments and oceanographic changes. In particular, foraminiferal pore traits are promising proxies for past ocean oxygenation. However, conventional 2D measurements from limited shell regions may obscure chamber-wise and ontogenetic variation. We present a deep learning and statistical pipeline for chamber-resolved analysis of pore traits in 3D micro-computed tomography (μCT) scans of foraminiferal shells, and demonstrate its use on the benthic species \textit{Elphidium clavatum}. The pipeline combines browser-based annotation, multi-planar 2D U-Net... (More)
- Foraminiferal shells are abundant in marine sediments, well preserved, and morphologically sensitive to environmental conditions, making them widely used archives of past marine environments and oceanographic changes. In particular, foraminiferal pore traits are promising proxies for past ocean oxygenation. However, conventional 2D measurements from limited shell regions may obscure chamber-wise and ontogenetic variation. We present a deep learning and statistical pipeline for chamber-resolved analysis of pore traits in 3D micro-computed tomography (μCT) scans of foraminiferal shells, and demonstrate its use on the benthic species \textit{Elphidium clavatum}. The pipeline combines browser-based annotation, multi-planar 2D U-Net segmentation, geometric post-processing, unsupervised pore clustering, and interactive correction of chamber-wise pore assignments. This framework enables volumetric porosity estimation, per-pore geometric characterization, pore-size distribution analysis, and chamber-wise summaries of pore traits. Applied to Baltic Sea specimens, the pipeline revealed chamber-wise pore variation across the final whorl: older chambers contained fewer but larger and more elongated pores, while pore thickness changed only weakly. This suggests that whole-shell summaries may mask biologically relevant chamber-specific variation. The proposed workflow provides a practical tool for linking 3D foraminiferal morphology to paleoenvironmental interpretations. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/3d4e190f-5548-4eb6-a36a-ceb11f35a541
- author
- Wu, Hanqing
LU
; Choquel, Constance
; Ni, Sha
; Filipsson, Helena L.
LU
and Pirzamanbin, Behnaz
LU
- organization
- publishing date
- 2026
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- in press
- subject
- keywords
- foraminifera, machine learning (ML), marine sciences
- host publication
- Computer Vision – ECCV 2026 : 19th European Conference, Malmö, Sweden, September 8–12, 2026, Proceedings, Part I - 19th European Conference, Malmö, Sweden, September 8–12, 2026, Proceedings, Part I
- series title
- Lecture Notes in Computer Science
- volume
- 1700
- publisher
- Springer
- ISSN
- 1611-3349
- project
- From Microns to Models: Leveraging µCT and Machine Learning for High-Resolution Morphological Diversity Analysis (MICROMORPH)
- A big data approach to environmental change: Statistical quantification of 3D microfossil images
- language
- English
- LU publication?
- yes
- id
- 3d4e190f-5548-4eb6-a36a-ceb11f35a541
- alternative location
- https://link.springer.com/book/9783032369833
- https://openreview.net/pdf?id=VGTTYP7CVX
- date added to LUP
- 2026-09-10 15:24:53
- date last changed
- 2026-09-10 16:32:04
@inproceedings{3d4e190f-5548-4eb6-a36a-ceb11f35a541,
abstract = {{Foraminiferal shells are abundant in marine sediments, well preserved, and morphologically sensitive to environmental conditions, making them widely used archives of past marine environments and oceanographic changes. In particular, foraminiferal pore traits are promising proxies for past ocean oxygenation. However, conventional 2D measurements from limited shell regions may obscure chamber-wise and ontogenetic variation. We present a deep learning and statistical pipeline for chamber-resolved analysis of pore traits in 3D micro-computed tomography (μCT) scans of foraminiferal shells, and demonstrate its use on the benthic species \textit{Elphidium clavatum}. The pipeline combines browser-based annotation, multi-planar 2D U-Net segmentation, geometric post-processing, unsupervised pore clustering, and interactive correction of chamber-wise pore assignments. This framework enables volumetric porosity estimation, per-pore geometric characterization, pore-size distribution analysis, and chamber-wise summaries of pore traits. Applied to Baltic Sea specimens, the pipeline revealed chamber-wise pore variation across the final whorl: older chambers contained fewer but larger and more elongated pores, while pore thickness changed only weakly. This suggests that whole-shell summaries may mask biologically relevant chamber-specific variation. The proposed workflow provides a practical tool for linking 3D foraminiferal morphology to paleoenvironmental interpretations.}},
author = {{Wu, Hanqing and Choquel, Constance and Ni, Sha and Filipsson, Helena L. and Pirzamanbin, Behnaz}},
booktitle = {{Computer Vision – ECCV 2026 : 19th European Conference, Malmö, Sweden, September 8–12, 2026, Proceedings, Part I}},
issn = {{1611-3349}},
keywords = {{foraminifera; machine learning (ML); marine sciences}},
language = {{eng}},
publisher = {{Springer}},
series = {{Lecture Notes in Computer Science}},
title = {{A Pipeline for Chamber-Resolved Analysis of Pore Traits in Foraminiferal µCT Volumes}},
url = {{https://link.springer.com/book/9783032369833}},
volume = {{1700}},
year = {{2026}},
}