Asynchronous Decentralized Federated Lifelong Learning for Landmark Localization in Medical Imaging

TL;DR

The paper addresses the challenge of accurate landmark localization in medical imaging under privacy constraints while requiring continual, multi-task learning. It introduces a decentralized, asynchronous federated lifelong learning (ADFLL) framework with hub-based communication and selective experience replay, enabling learning from diverse environments without a central server. Empirical results on the BraTS dataset show that ADFLL outperforms all-knowing and traditional lifelong RL baselines, with strong scalability and robustness to network dropout and agent churn. This approach offers a privacy-preserving, scalable solution for continual medical image analysis across heterogeneous devices and imaging environments, with potential for real-world clinical deployment.

Abstract

Federated learning is a recent development in the machine learning area that allows a system of devices to train on one or more tasks without sharing their data to a single location or device. However, this framework still requires a centralized global model to consolidate individual models into one, and the devices train synchronously, which both can be potential bottlenecks for using federated learning. In this paper, we propose a novel method of asynchronous decentralized federated lifelong learning (ADFLL) method that inherits the merits of federated learning and can train on multiple tasks simultaneously without the need for a central node or synchronous training. Thus, overcoming the potential drawbacks of conventional federated learning. We demonstrate excellent performance on the brain tumor segmentation (BRATS) dataset for localizing the left ventricle on multiple image sequences and image orientation. Our framework allows agents to achieve the best performance with a mean distance error of 7.81, better than the conventional all-knowing agent's mean distance error of 11.78, and significantly (p=0.01) better than a conventional lifelong learning agent with a distance error of 15.17 after eight rounds of training. In addition, all ADFLL agents have comparable or better performance than a conventional LL agent. In conclusion, we developed an ADFLL framework with excellent performance and speed-up compared to conventional RL agents.

Figures (7)

• Figure 1: Illustration of the 8 task-environment pairs. The red boxes indicate the true landmark location of the top left ventricle. The yellow box is a predicted location from ADFLL agents during their training progressions
• Figure 2: Illustration of the 4-agent decentralized federated lifelong learning framework of our experiment.
• Figure 3: Comparison of distance error across 8 different tasks between Agent X (Central Aggregation) and Agent 2 (ADFLL).
• Figure 4: Comparison of distance error of all agents in the system across 4 rounds of training
• Figure 5: Comparison of distance error of all agents in the system across 4 rounds of training