TY - GEN
T1 - On Differentially Private 3D Medical Image Synthesis with Controllable Latent Diffusion Models
AU - Daum, Deniz
AU - Osuala, Richard
AU - Riess, Anneliese
AU - Kaissis, Georgios
AU - Schnabel, Julia A.
AU - Di Folco, Maxime
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.
PY - 2025
Y1 - 2025
N2 - Generally, the small size of public medical imaging datasets coupled with stringent privacy concerns, hampers the advancement of data-hungry deep learning models in medical imaging. This study addresses these challenges for 3D cardiac MRI images in the short-axis view. We propose Latent Diffusion Models that generate synthetic images conditioned on medical attributes, while ensuring patient privacy through differentially private model training. To our knowledge, this is the first work to apply and quantify differential privacy in 3D medical image generation. We pre-train our models on public data and finetune them with differential privacy on the UK Biobank dataset. Our experiments reveal that pre-training significantly improves model performance, achieving a Fréchet Inception Distance (FID) of 26.77 at ϵ=10, compared to 92.52 for models without pre-training. Additionally, we explore the trade-off between privacy constraints and image quality, investigating how tighter privacy budgets affect output controllability and may lead to degraded performance. Our results demonstrate that proper consideration during training with differential privacy can substantially improve the quality of synthetic cardiac MRI images, but there are still notable challenges in achieving consistent medical realism. Code: https://github.com/compai-lab/2024-miccai-dgm-daum.
AB - Generally, the small size of public medical imaging datasets coupled with stringent privacy concerns, hampers the advancement of data-hungry deep learning models in medical imaging. This study addresses these challenges for 3D cardiac MRI images in the short-axis view. We propose Latent Diffusion Models that generate synthetic images conditioned on medical attributes, while ensuring patient privacy through differentially private model training. To our knowledge, this is the first work to apply and quantify differential privacy in 3D medical image generation. We pre-train our models on public data and finetune them with differential privacy on the UK Biobank dataset. Our experiments reveal that pre-training significantly improves model performance, achieving a Fréchet Inception Distance (FID) of 26.77 at ϵ=10, compared to 92.52 for models without pre-training. Additionally, we explore the trade-off between privacy constraints and image quality, investigating how tighter privacy budgets affect output controllability and may lead to degraded performance. Our results demonstrate that proper consideration during training with differential privacy can substantially improve the quality of synthetic cardiac MRI images, but there are still notable challenges in achieving consistent medical realism. Code: https://github.com/compai-lab/2024-miccai-dgm-daum.
KW - Cardiac MRI
KW - Differential Privacy
KW - Generative Models
KW - Synthetic Data
UR - http://www.scopus.com/inward/record.url?scp=85206989618&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-72744-3_14
DO - 10.1007/978-3-031-72744-3_14
M3 - Conference contribution
AN - SCOPUS:85206989618
SN - 9783031727436
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 139
EP - 149
BT - Deep Generative Models - 4th MICCAI Workshop, DGM4MICCAI 2024, Held in Conjunction with MICCAI 2024, Proceedings
A2 - Mukhopadhyay, Anirban
A2 - Oksuz, Ilkay
A2 - Engelhardt, Sandy
A2 - Mehrof, Dorit
A2 - Yuan, Yixuan
PB - Springer Science and Business Media Deutschland GmbH
T2 - 4th Workshop on Deep Generative Models for Medical Image Computing and Computer Assisted Intervention, DGM4MICCAI 2024, held in Conjunction with 27th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2024
Y2 - 10 October 2024 through 10 October 2024
ER -