FedDermaSeg: Federated Learning for Dermatological Image Segmentation

Anabik Pal; Ganesh Patidar; Bikash Santra

cs.CVOct 6, 2026

FedDermaSeg: Federated Learning for Dermatological Image Segmentation

Authors: Anabik Pal, Ganesh Patidar, Bikash Santra

Organizations: Department of Computer Science, Indian Institute of Science Education and Research Berhampur, India · Department of Mathematics, Indian Institute of Technology Jodhpur, India · School of Artificial Intelligence and Data Science, Indian Institute of Technology Jodhpur, India

Abstract

Skin cancer is a major global health concern, and early detection and accurate lesion delineation are important for effective diagnosis and treatment planning. Automated skin lesion analysis can assist dermatologists, with lesion segmentation serving as a fundamental step in computer-aided diagnostic systems. Conventional deep learning-based segmentation models typically rely on centralized training, where images and their corresponding segmentation masks are collected on a central server. Such data aggregation raises privacy concerns in medical applications and requires substantial centralized computational resources. To address these limitations, we investigate the feasibility of federated learning for privacy-preserving skin lesion segmentation. The training and validation sets of the ISIC 2018 Skin Lesion Segmentation Challenge dataset are used to simulate a distributed learning environment and develop a federated segmentation model. The resulting model is evaluated on the ISIC 2018 test set and the PH2 dataset to assess its performance and generalizability. Experimental results demonstrate that the federated model achieves performance comparable to centralized training while consistently improving upon the locally trained models. These findings demonstrate the potential of federated learning for collaborative skin lesion segmentation without requiring centralized aggregation of medical images.

Figures & tables

Figure 1: The flow diagram of training segmentation network with federated learning. $Org_{i}$ refers to $i^{th}$ client; $W\_Orgi$ refers to local weight update in $i^{th}$ client during a communication round; $W\_Fed$ refers to global model after local model aggregation.

Network	Model	Dice	IoU	Accuracy	Precision	Recall	F1
UNet	local-1	0.882	0.789	0.937	0.934	0.835	0.882
	local-2	0.883	0.791	0.936	0.92	0.849	0.883
	local-3	0.880	0.786	0.935	0.912	0.850	0.880
	local-4	0.877	0.782	0.933	0.909	0.848	0.877
	local-5	0.881	0.787	0.935	0.911	0.853	0.881
	global-L	0.919	0.850	0.950	0.908	0.930	0.919

Table 1: Performance in the test set of ISIC 2018 skin lesion segmentation challenge dataset

Network	Model	Dice	IoU	Accuracy	Precision	Recall	F1
UNet	local-1	0.854	0.745	0.911	0.905	0.808	0.854
	local-2	0.851	0.741	0.907	0.88	0.824	0.851
	local-3	0.858	0.751	0.911	0.887	0.831	0.858
	local-4	0.847	0.735	0.903	0.861	0.833	0.847
	local-5	0.846	0.733	0.905	0.888	0.808	0.846
	global-L	0.869	0.769	0.915	0.862	0.877	0.869

Table 2: Results of various models trained with ISIC 2018 and tested with the

PH^{2}

dataset

Figure 2: Pictorial representation of segmentation results generated by the global-L and global-FL models for both TransUNet and UNet tested with ISIC 2018 dataset

Figure 4: Performances yield by the global-FL TransUNet model trained and tested on the ISIC 2018 dataset, when the #epochs in each iteration during federated learning is varied from 1 (epoch_1) to 5 (epoch_5) and 10 (epoch_10).

Explore similar work

Similarity Weighted Aggregation with Global Differential Privacy for Federated Brain Lesion Segmentation
Aug 1, 2026Muhammad Irfan Khan, Eero Lehtonen, Joni Obradovic +4Federated Learning Brain Tumor Segmentation
Turku University of Applied Sciences, Turku 20520, Finland
QFedPolyp: A Communication- and Inference-Efficient Federated Learning Framework for Polyp Segmentation
Jul 23, 2026Madan Baduwal, Priyanka PaudelPolyp Segmentation Federated Learning
Department of Computer Science and Engineering, Mississippi State University Mississippi State, MS, USA
MuCALD-SplitFed: Causal-Latent Diffusion for Privacy-Preserving Multi-Task Split-Federated Medical Image Segmentation
May 4, 2026Chamani Shiranthika, Hadi Hadizadeh, Parvaneh SaeediSemi-Supervised Medical Image Segmentation Identity-Conditioned Latent Diffusion Models
School of Engineering Science, Simon Fraser University, Burnaby, BC, Canada

Aug 1, 2026, cs.AIJ/K move · Enter open · S save

Similarity Weighted Aggregation with Global Differential Privacy for Federated Brain Lesion Segmentation

Muhammad Irfan Khan, Eero Lehtonen, Joni Obradovic, Elina Kontio+3

Turku University of Applied Sciences, Turku 20520, Finland

Federated Learning Brain Tumor Segmentation Δ)$-Differential Privacy Magnetic Resonance Imaging Privacy Nvidia Isaac Sim+1

Open paper AskPDF arXiv

Federated Learning (FL) enables collaborative training of machine learning models across multiple institutions without sharing sensitive data, making it particularly suitable for medical imaging applications. However, heterogeneous data distributions across institutions and potential information leakage through model updates remain important challenges. In this work, we propose DP-SimAgg, a privacy-preserving federated learning framework that integrates similarity-weighted aggregation with a server-side differential privacy mechanism. The proposed method applies L2 clipping to bound collaborator updates, computes similarity-based aggregation weights to mitigate the effects of non-IID data distributions, and injects calibrated Gaussian noise at the central server, providing per-round privacy guarantees under the assumed sensitivity bound. The framework is implemented using Intel's OpenFL platform and evaluated on the FeTS 2022 dataset consisting of 1251 multi-modal MRI scans for brain tumor segmentation. Experimental results demonstrate that DP-SimAgg maintains competitive segmentation performance while providing privacy protection. Under a strict per-round privacy budget (epsilon = 1, cumulative epsilon_total = 20 over 20 rounds), the method achieves Dice scores of 0.6357, 0.5305, and 0.5274 for the enhancing tumor (ET), tumor core (TC), and whole tumor (WT) regions, respectively. With a more relaxed per-round budget (epsilon = 10, cumulative epsilon_total = 200), performance approaches that of the non-private baseline while incorporating a central Gaussian mechanism with per-round (epsilon, delta)-DP accounting under the assumed sensitivity bound. These results highlight the potential of DP-SimAgg for enabling privacy-preserving collaborative learning in medical imaging applications.

Abstract

Explore similar work

Similarity Weighted Aggregation with Global Differential Privacy for Federated Brain Lesion Segmentation

QFedPolyp: A Communication- and Inference-Efficient Federated Learning Framework for Polyp Segmentation

MuCALD-SplitFed: Causal-Latent Diffusion for Privacy-Preserving Multi-Task Split-Federated Medical Image Segmentation