Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification

Kaouther Mouheb; Marawan Elbatel; Janne Papma; Geert Jan Biessels; Jurgen Claassen; Huub Middelkoop; Barbara van Munster; Wiesje van der Flier; Inez Ramakers; Stefan Klein; Esther E. Bron; G Zamzmi; A Reinke; R Samala; M Jiang; X Li; H Roth; M Sidulova; T Kooi; S Albarqouni; S Bakas; N Rieke

doi:10.1007/978-3-032-05663-4_7

Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification

Kaouther Mouheb^*
, Marawan Elbatel
, Janne Papma
, Geert Jan Biessels
, Jurgen Claassen
, Huub Middelkoop
, Barbara van Munster
, Wiesje van der Flier
, Inez Ramakers
, Stefan Klein
, Esther E. Bron
, G Zamzmi
, A Reinke
, R Samala
, M Jiang
, X Li
, H Roth
, M Sidulova
, T Kooi
, S Albarqouni

S Bakas, N Rieke

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

Abstract

While foundation models (FMs) offer strong potential for AI-based dementia diagnosis, their integration into federated learning (FL) systems remains underexplored. In this benchmarking study, we systematically evaluate the impact of key design choices: classification head architecture, fine-tuning strategy, and aggregation method, on the performance and efficiency of federated FM tuning using brain MRI data. Using a large multi-cohort dataset, we find that the architecture of the classification head substantially influences performance, freezing the FM encoder achieves comparable results to full fine-tuning, and advanced aggregation methods outperform standard federated averaging. Our results offer practical insights for deploying FMs in decentralized clinical settings and highlight trade-offs that should guide future method development.

Original language	English
Title of host publication	Bridging Regulatory Science and Medical Imaging Evaluation; and Distributed, Collaborative, and Federated Learning - 1st International Workshop, BRIDGE 2025, and 6th International Workshop, DeCaF 2025, Held in Conjunction with MICCAI 2025, Proceedings
Editors	Ghada Zamzmi, Annika Reinke, Ravi Samala, Meirui Jiang, Xiaoxiao Li, Holger Roth, Mariia Sidulova, Thijs Kooi, Shadi Albarqouni, Spyridon Bakas, Nicola Rieke
Publisher	Springer
Pages	69-79
Number of pages	11
ISBN (Electronic)	9783032056658
ISBN (Print)	9783032056627
DOIs	https://doi.org/10.1007/978-3-032-05663-4_7
Publication status	Published - 2026
Event	2025 International Workshop on Bridging Regulatory Science and Medical Imaging Evaluation-BRIDGE - Daejeon, Korea, Republic of Duration: 23 Sept 2025 → 23 Sept 2025 https://bridge-regsci.github.io/miccaibridge/home/

Publication series

Series	Lecture Notes in Computer Science
Volume	16135
ISSN	0302-9743

Workshop

Workshop	2025 International Workshop on Bridging Regulatory Science and Medical Imaging Evaluation-BRIDGE
Country/Territory	Korea, Republic of
City	Daejeon
Period	23/09/25 → 23/09/25
Internet address	https://bridge-regsci.github.io/miccaibridge/home/

Keywords

Federated learning
Foundation models
Dementia
MRI

Access to Document

10.1007/978-3-032-05663-4_7

https://pure.amsterdamumc.nl/en/publications/13bfff0f-507b-4b6d-998b-3406d5c2e3adLicence: CC BY

Cite this

Mouheb, K., Elbatel, M., Papma, J., Biessels, G. J., Claassen, J., Middelkoop, H., van Munster, B., van der Flier, W., Ramakers, I., Klein, S., Bron, E. E., Zamzmi, G., Reinke, A., Samala, R., Jiang, M., Li, X., Roth, H., Sidulova, M., Kooi, T., ... Rieke, N. (2026). Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification. In G. Zamzmi, A. Reinke, R. Samala, M. Jiang, X. Li, H. Roth, M. Sidulova, T. Kooi, S. Albarqouni, S. Bakas, & N. Rieke (Eds.), Bridging Regulatory Science and Medical Imaging Evaluation; and Distributed, Collaborative, and Federated Learning - 1st International Workshop, BRIDGE 2025, and 6th International Workshop, DeCaF 2025, Held in Conjunction with MICCAI 2025, Proceedings (pp. 69-79). Springer. https://doi.org/10.1007/978-3-032-05663-4_7

Mouheb, Kaouther ; Elbatel, Marawan ; Papma, Janne et al. / Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification. Bridging Regulatory Science and Medical Imaging Evaluation; and Distributed, Collaborative, and Federated Learning - 1st International Workshop, BRIDGE 2025, and 6th International Workshop, DeCaF 2025, Held in Conjunction with MICCAI 2025, Proceedings. editor / Ghada Zamzmi ; Annika Reinke ; Ravi Samala ; Meirui Jiang ; Xiaoxiao Li ; Holger Roth ; Mariia Sidulova ; Thijs Kooi ; Shadi Albarqouni ; Spyridon Bakas ; Nicola Rieke. Springer, 2026. pp. 69-79 (Lecture Notes in Computer Science, Vol. 16135).

@inproceedings{e0cc93d43d114c89b3fe0682a3a7a9eb,

title = "Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification",

abstract = "While foundation models (FMs) offer strong potential for AI-based dementia diagnosis, their integration into federated learning (FL) systems remains underexplored. In this benchmarking study, we systematically evaluate the impact of key design choices: classification head architecture, fine-tuning strategy, and aggregation method, on the performance and efficiency of federated FM tuning using brain MRI data. Using a large multi-cohort dataset, we find that the architecture of the classification head substantially influences performance, freezing the FM encoder achieves comparable results to full fine-tuning, and advanced aggregation methods outperform standard federated averaging. Our results offer practical insights for deploying FMs in decentralized clinical settings and highlight trade-offs that should guide future method development.",

keywords = "Federated learning, Foundation models, Dementia, MRI",

author = "Kaouther Mouheb and Marawan Elbatel and Janne Papma and Biessels, \{Geert Jan\} and Jurgen Claassen and Huub Middelkoop and \{van Munster\}, Barbara and \{van der Flier\}, Wiesje and Inez Ramakers and Stefan Klein and Bron, \{Esther E.\} and G Zamzmi and A Reinke and R Samala and M Jiang and X Li and H Roth and M Sidulova and T Kooi and S Albarqouni and S Bakas and N Rieke",

year = "2026",

doi = "10.1007/978-3-032-05663-4\_7",

language = "English",

isbn = "9783032056627",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "69--79",

editor = "Ghada Zamzmi and Annika Reinke and Ravi Samala and Meirui Jiang and Xiaoxiao Li and Holger Roth and Mariia Sidulova and Thijs Kooi and Shadi Albarqouni and Spyridon Bakas and Nicola Rieke",

booktitle = "Bridging Regulatory Science and Medical Imaging Evaluation; and Distributed, Collaborative, and Federated Learning - 1st International Workshop, BRIDGE 2025, and 6th International Workshop, DeCaF 2025, Held in Conjunction with MICCAI 2025, Proceedings",

address = "United States",

note = "2025 International Workshop on Bridging Regulatory Science and Medical Imaging Evaluation-BRIDGE ; Conference date: 23-09-2025 Through 23-09-2025",

url = "https://bridge-regsci.github.io/miccaibridge/home/",

}

Mouheb, K, Elbatel, M, Papma, J, Biessels, GJ, Claassen, J, Middelkoop, H, van Munster, B, van der Flier, W, Ramakers, I, Klein, S, Bron, EE, Zamzmi, G, Reinke, A, Samala, R, Jiang, M, Li, X, Roth, H, Sidulova, M, Kooi, T, Albarqouni, S, Bakas, S & Rieke, N 2026, Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification. in G Zamzmi, A Reinke, R Samala, M Jiang, X Li, H Roth, M Sidulova, T Kooi, S Albarqouni, S Bakas & N Rieke (eds), Bridging Regulatory Science and Medical Imaging Evaluation; and Distributed, Collaborative, and Federated Learning - 1st International Workshop, BRIDGE 2025, and 6th International Workshop, DeCaF 2025, Held in Conjunction with MICCAI 2025, Proceedings. Springer, Lecture Notes in Computer Science, vol. 16135, pp. 69-79, 2025 International Workshop on Bridging Regulatory Science and Medical Imaging Evaluation-BRIDGE, Daejeon, Korea, Republic of, 23/09/25. https://doi.org/10.1007/978-3-032-05663-4_7

Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification. / Mouheb, Kaouther; Elbatel, Marawan; Papma, Janne et al.
Bridging Regulatory Science and Medical Imaging Evaluation; and Distributed, Collaborative, and Federated Learning - 1st International Workshop, BRIDGE 2025, and 6th International Workshop, DeCaF 2025, Held in Conjunction with MICCAI 2025, Proceedings. ed. / Ghada Zamzmi; Annika Reinke; Ravi Samala; Meirui Jiang; Xiaoxiao Li; Holger Roth; Mariia Sidulova; Thijs Kooi; Shadi Albarqouni; Spyridon Bakas; Nicola Rieke. Springer, 2026. p. 69-79 (Lecture Notes in Computer Science, Vol. 16135).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

TY - GEN

T1 - Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification

AU - Mouheb, Kaouther

AU - Elbatel, Marawan

AU - Papma, Janne

AU - Biessels, Geert Jan

AU - Claassen, Jurgen

AU - Middelkoop, Huub

AU - van Munster, Barbara

AU - van der Flier, Wiesje

AU - Ramakers, Inez

AU - Klein, Stefan

AU - Bron, Esther E.

AU - Zamzmi, G

AU - Reinke, A

AU - Samala, R

AU - Jiang, M

AU - Li, X

AU - Roth, H

AU - Sidulova, M

AU - Kooi, T

AU - Albarqouni, S

AU - Bakas, S

AU - Rieke, N

PY - 2026

Y1 - 2026

N2 - While foundation models (FMs) offer strong potential for AI-based dementia diagnosis, their integration into federated learning (FL) systems remains underexplored. In this benchmarking study, we systematically evaluate the impact of key design choices: classification head architecture, fine-tuning strategy, and aggregation method, on the performance and efficiency of federated FM tuning using brain MRI data. Using a large multi-cohort dataset, we find that the architecture of the classification head substantially influences performance, freezing the FM encoder achieves comparable results to full fine-tuning, and advanced aggregation methods outperform standard federated averaging. Our results offer practical insights for deploying FMs in decentralized clinical settings and highlight trade-offs that should guide future method development.

AB - While foundation models (FMs) offer strong potential for AI-based dementia diagnosis, their integration into federated learning (FL) systems remains underexplored. In this benchmarking study, we systematically evaluate the impact of key design choices: classification head architecture, fine-tuning strategy, and aggregation method, on the performance and efficiency of federated FM tuning using brain MRI data. Using a large multi-cohort dataset, we find that the architecture of the classification head substantially influences performance, freezing the FM encoder achieves comparable results to full fine-tuning, and advanced aggregation methods outperform standard federated averaging. Our results offer practical insights for deploying FMs in decentralized clinical settings and highlight trade-offs that should guide future method development.

KW - Federated learning

KW - Foundation models

KW - Dementia

KW - MRI

U2 - 10.1007/978-3-032-05663-4_7

DO - 10.1007/978-3-032-05663-4_7

M3 - Conference article in proceeding

SN - 9783032056627

T3 - Lecture Notes in Computer Science

SP - 69

EP - 79

BT - Bridging Regulatory Science and Medical Imaging Evaluation; and Distributed, Collaborative, and Federated Learning - 1st International Workshop, BRIDGE 2025, and 6th International Workshop, DeCaF 2025, Held in Conjunction with MICCAI 2025, Proceedings

A2 - Zamzmi, Ghada

A2 - Reinke, Annika

A2 - Samala, Ravi

A2 - Jiang, Meirui

A2 - Li, Xiaoxiao

A2 - Roth, Holger

A2 - Sidulova, Mariia

A2 - Kooi, Thijs

A2 - Albarqouni, Shadi

A2 - Bakas, Spyridon

A2 - Rieke, Nicola

PB - Springer

T2 - 2025 International Workshop on Bridging Regulatory Science and Medical Imaging Evaluation-BRIDGE

Y2 - 23 September 2025 through 23 September 2025

ER -

Mouheb K, Elbatel M, Papma J, Biessels GJ, Claassen J, Middelkoop H et al. Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification. In Zamzmi G, Reinke A, Samala R, Jiang M, Li X, Roth H, Sidulova M, Kooi T, Albarqouni S, Bakas S, Rieke N, editors, Bridging Regulatory Science and Medical Imaging Evaluation; and Distributed, Collaborative, and Federated Learning - 1st International Workshop, BRIDGE 2025, and 6th International Workshop, DeCaF 2025, Held in Conjunction with MICCAI 2025, Proceedings. Springer. 2026. p. 69-79. (Lecture Notes in Computer Science, Vol. 16135). doi: 10.1007/978-3-032-05663-4_7

Federated Fine-Tuning of SAM-Med3D for MRI-Based Dementia Classification

Abstract

Publication series

Workshop

Keywords

Access to Document

Fingerprint

Cite this