Deep Supervision Attention U-net for segmentation of uterine zones: a multi-center study

Smiti Tripathy, Nyvenn Castro, N. Castro, Matthias May, Lisa Siegler, Lisa Story, Michael Uder, Jana Hutter
In: ISMRM Annual Meeting · 2025 · doi:10.58530/2025/0105 · W4414237272
article OA: closed CC0
Full text JSON View on OpenAlex View at publisher
⚙ AI-generated summary by gemini-2.5-flash-lite, 2026-06-09 ⓘ

This study developed a deeply supervised Attention U-net for automated segmentation of uterine zones, achieving a mean Dice score of 0.8178 on a multi-center dataset.

One-sentence paraphrase of the abstract; not a substitute for reading it. No clinical advice. How this works

⚙ AI-generated deep summary by claude@2026-06, 2026-06-18 · read from full text ⓘ

This multi-center study investigated whether a deeply supervised Attention-based 3D U-net could automatically segment uterine layers (uterine zones) from MRI scans in a way that is robust across different acquisition specifics. Using a multicenter, multi–field strength dataset, the authors trained the network on uterine-layer segmentation and evaluated performance using overlap metrics. They reported a mean Dice score of 0.8178 and a mean Jaccard index of 0.7176. The paper notes that while MRI has superior soft-tissue contrast, automated assessment options are lacking, and frames the limitation as the need for automatic quantification of uterine microstructure for downstream analysis. This paper is centrally about endometriosis and/or adenomyosis — it directly motivates automated uterine-layer quantification because uterine wall microstructure is altered in adenomyosis, and it is positioned to facilitate future study of adenomyosis using the segmented uterine zones.

Read from the paper's body, not the abstract. Not a substitute for reading the paper. No clinical advice. How this works

Abstract

Motivation: The microstructure in the wall of the uterus is key for normal uterine physiology throughout the menstrual cycle and altered in diseases such as adenomyosis. Visualization and quantification in-vivo is thus essential. While MRI offers superior soft tissue contrast, the analysis is limited by a lack of automatic assessing options. Goal(s): Robust automatic segmentation of uterine layers, independent of acquisition specifics. Approach: Deeply supervised Attention-based 3D U-net trained on a multicenter multi field strength dataset for automated segmentation of uterine layers. Results: The proposed network achieved a mean Dice score of 0.8178 and a mean Jaccard index of 0.7176. Impact: This study enabled the deep learning-based automatic segmentation of the uterine zones, that would in future provide deeper insights into uterine layer changes at different menstrual cycle points and facilitate the study of Adenomyosis and other uterine abnormalities.
Full text 2,437 characters · extracted from oa-doi-fallback · 3 sections · click to expand

Abstract

#0105 Deep Supervision Attention U-net for segmentation of uterine zones: a multi-center study Smiti Tripathy1,2, Nyvenn Castro1,2, Matthias May2, Lisa Siegler2, Lisa Story3,4,5, Michael Uder2, and Jana Hutter1,2,3,4 1Smart Imaging Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany, 2Radiologisches Institut, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany, 3Biomedical Imaging Department, School of Biomedical Engineering and Imaging, KCL, London, United Kingdom, 4Early Life Imaging Department, School of Biomedical Engineering and Imaging, KCL, London, United Kingdom, 5Women’s Health Academic Department, KCL, London, United Kingdom Synopsis

Keywords

Uterus, Uterus, Automatic Segmentation Motivation: The microstructure in the wall of the uterus is key for normal uterine physiology throughout the menstrual cycle and altered in diseases such as adenomyosis. Visualization and quantification in-vivo is thus essential. While MRI offers superior soft tissue contrast, the analysis is limited by a lack of automatic assessing options. Goal(s): Robust automatic segmentation of uterine layers, independent of acquisition specifics. Approach: Deeply supervised Attention-based 3D U-net trained on a multicenter multi field strength dataset for automated segmentation of uterine layers.

Results

The proposed network achieved a mean Dice score of 0.8178 and a mean Jaccard index of 0.7176. Impact: This study enabled the deep learning-based automatic segmentation of the uterine zones, that would in future provide deeper insights into uterine layer changes at different menstrual cycle points and facilitate the study of Adenomyosis and other uterine abnormalities. How to access this content: For one year after publication, abstracts and videos are only open to registrants of this annual meeting. Registrants should use their existing login information. Non-registrant access can be purchased via the ISMRM E-Library. After one year, current ISMRM & ISMRT members get free access to both the abstracts and videos. Non-members and non-registrants must purchase access via the ISMRM E-Library. After two years, the meeting proceedings (abstracts) are opened to the public and require no login information. Videos remain behind password for access by members, registrants and E-Library customers. Click here for more information on becoming a member.

Text is read by the "Ask this paper" AI Q&A widget below. Extraction quality varies by source — PMC NXML preserves structure cleanly, OA-HTML may include some navigation residue, and OA-PDF can have broken hyphenation. The publisher copy (via DOI) is the canonical version.

My notes (saved in your browser only)

⚙ Ask this paper AI returns verbatim quotes from the full text · source: oa-doi-fallback ⓘ

Answers must be backed by verbatim quotes from this paper's full text. Hallucinated quotes are dropped automatically; if no verbatim passage answers the question, we say so. How this works

Condition tags

adenomyosis

Citation neighborhood (no data yet)

We don't have any in-corpus citations linked to this paper yet. This is a recent paper (2025) — citers typically take a year or two to land, and the OpenAlex reference graph may still be filling in.

Source provenance

openalex
last seen: 2026-06-04T00:00:01.174412+00:00
License: CC0 · commercial use OK