[Paper Review] Deep Anatomical Federated Network (Dafne): An open client-server framework for the continuous, collaborative improvement of deep learning-based medical image segmentation

Francesco Santini, Jakob Wasserthal+12 more|arXiv (Cornell University)|Feb 13, 2023

Radiomics and Machine Learning in Medical Imaging20 citations

TL;DR

Dafne is a free, decentralized client-server system for federated incremental learning to improve medical image segmentation, validated on MRI data with measurable performance gains over generations.

ABSTRACT

Purpose: To present and evaluate Dafne (deep anatomical federated network), a freely available decentralized, collaborative deep learning system for the semantic segmentation of radiological images through federated incremental learning. Materials and Methods: Dafne is free software with a client-server architecture. The client side is an advanced user interface that applies the deep learning models stored on the server to the user's data and allows the user to check and refine the prediction. Incremental learning is then performed at the client's side and sent back to the server, where it is integrated into the root model. Dafne was evaluated locally, by assessing the performance gain across model generations on 38 MRI datasets of the lower legs, and through the analysis of real-world usage statistics (n = 639 use-cases). Results: Dafne demonstrated a statistically improvement in the accuracy of semantic segmentation over time (average increase of the Dice Similarity Coefficient by 0.007 points/generation on the local validation set, p < 0.001). Qualitatively, the models showed enhanced performance on various radiologic image types, including those not present in the initial training sets, indicating good model generalizability. Conclusion: Dafne showed improvement in segmentation quality over time, demonstrating potential for learning and generalization.

Motivation & Objective

• Motivate and enable collaborative, continuous improvement of DL-based medical image segmentation without centralized data sharing.
• Provide a freely available client-server architecture that supports incremental learning from user refinements.
• Evaluate the approach on real-world MRI datasets to assess generalization and learning over time.

Proposed method

• Dafne uses a client-server architecture with a server hosting the root segmentation model and clients applying models to local data.
• Clients refine predictions and perform incremental learning locally, sending updates back to the server for integration into the root model.
• The system enables continuous, collaborative model improvement through federated incremental learning.
• Evaluation includes local performance gains across model generations on MRI data and analysis of real-world usage statistics.

Experimental results

Research questions

• RQ1 Can decentralized, incremental learning via Dafne improve segmentation accuracy over successive model generations?
• RQ2 Do client-side refinements and federated integration enhance generalization to radiologic images not in the initial training set?
• RQ3 What is the practical impact of using Dafne on real-world usage (usage statistics, scalability)?

Key findings

• Dafne showed a statistically significant improvement in segmentation accuracy over time, with an average Dice Similarity Coefficient increase of 0.007 points per generation on the local validation set (p < 0.001).
• Qualitatively, models improved across diverse radiologic image types, including data not present in the initial training set, indicating good generalizability.
• Evaluation included 38 MRI datasets of the lower legs and 639 real-world use-cases, demonstrating practical applicability and learning potential.