Skip to search boxSkip to navigationSkip to main content

MedSymmFlow: Bridging Generative Modeling and Classification in Medical Imaging Through Symmetrical Flow Matching

Output:
Chapter in Book/Report/Conference proceeding
Conference contribution
Academic
Peer-review

Open access

Information

Output type

Output:
Chapter in Book/Report/Conference proceeding
Conference contribution
Academic
Peer-review

Host publication Subtitle

5th MICCAI Workshop, DGM4MICCAI 2025, Held in Conjunction with MICCAI 2025, Daejeon, South Korea, September 23, 2025, Proceedings

Original language

English

Pages from-to (Number of pages)

Pages 35-45 (11 pages)

Publication milestones

  • Published - 25/09/2025

Publication status

Published - 25/09/2025

Place of publication

Cham

Publisher

Springer, Netherlands, United States, United Kingdom, Germany, Italy, France, Austria, India, Switzerland, China, Japan

Publication series

  • Publication series name: Lecture Notes in Computer Science (LNCS)
    ISSN (Print): 0302-9743
    ISSN (Electronic): 1611-3349
    Volume: 16128
978-3-032-05471-5

ISBN (Electronic)

978-3-032-05472-2

Host publication title

Deep Generative Models

Host publication editors

  • Anirban Mukhopadhyay
  • Ilkay Oksuz
  • Sandy Engelhardt
  • Dorit Mehrof
  • Yixuan Yuan

Related Event

Title

5th Workshop on Deep Generative Models for Medical Image Computing and Computer Assisted Intervention, DGM4MICCAI 2025, held in conjunction with the 28th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2025

Event type

Conference

Date

23/09/2025 - 23/09/2025

Location

DaejeonKorea, Republic of

Abstract

<p>Reliable medical image classification requires accurate predictions and well-calibrated uncertainty estimates, especially in high-stakes clinical settings. This work presents MedSymmFlow, a generative-discriminative hybrid model built on Symmetrical Flow Matching, designed to unify classification, generation, and uncertainty quantification in medical imaging. MedSymmFlow leverages a latent-space formulation that scales to high-resolution inputs and introduces a semantic mask conditioning mechanism to enhance diagnostic relevance. Unlike standard discriminative models, it naturally estimates uncertainty through its generative sampling process. The model is evaluated on four MedMNIST datasets, covering a range of modalities and pathologies. The results show that MedSymmFlow matches or exceeds the performance of established baselines in classification accuracy and AUC, while also delivering reliable uncertainty estimates validated by performance improvements under selective prediction. The code is available at github.com/caetas/MedSymmFlow.</p>

Bibliographical Note

Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

Metrics

Metrics

SciVal
citations
1
Scopus
citations
1
Download statistics
Download count
15
SciVal
FWCI
3.70
SciVal
Author count
5
SciVal
Paper percentile
92
SciVal
Top percentile
10

PlumX, opens in new tab

Captures
6
Citation count
1