ADMarker: A Multi-Modal Federated Learning System for Monitoring Digital Biomarkers of Alzheimer's Disease

TL;DR

ADMarker tackles Alzheimer's disease monitoring by integrating multi-modal sensor data with a privacy-preserving three-stage federated learning framework to detect a broad set of digital biomarkers in home environments. By combining unsupervised and weakly supervised FL, the system handles limited labels, data heterogeneity, and constrained compute while keeping data private. In a four-week clinical deployment with 91 older adults, it achieved up to 93.8% biomarker detection accuracy and 88.9% accuracy for early AD diagnosis, demonstrating feasibility for longitudinal screening and personalized intervention. The work highlights the value of interpretable, multimodal biomarkers and federated privacy in scalable digital-health AI for neurodegenerative diseases.

Abstract

Alzheimer's Disease (AD) and related dementia are a growing global health challenge due to the aging population. In this paper, we present ADMarker, the first end-to-end system that integrates multi-modal sensors and new federated learning algorithms for detecting multidimensional AD digital biomarkers in natural living environments. ADMarker features a novel three-stage multi-modal federated learning architecture that can accurately detect digital biomarkers in a privacy-preserving manner. Our approach collectively addresses several major real-world challenges, such as limited data labels, data heterogeneity, and limited computing resources. We built a compact multi-modality hardware system and deployed it in a four-week clinical trial involving 91 elderly participants. The results indicate that ADMarker can accurately detect a comprehensive set of digital biomarkers with up to 93.8% accuracy and identify early AD with an average of 88.9% accuracy. ADMarker offers a new platform that can allow AD clinicians to characterize and track the complex correlation between multidimensional interpretable digital biomarkers, demographic factors of patients, and AD diagnosis in a longitudinal manner.

Figures (16)

• Figure 1: Overview of ADMarker. ADMarker consists of three major components, i.e., a multi-modal sensor system, federated learning for biomarker detection, and AD analysis based on detected digital biomarkers.
• Figure 2: Examples of data distributions of participants in a real-world deployment. The data distribution is highly imbalanced among different classes and non-i.i.d across different participants.
• Figure 3: Times of sensor down and bandwidth of all nodes over four weeks of the real-world deployment.
• Figure 4: The three-stage federated learning architecture of ADMarker. Stage 1: Centralized model pre-training; Stage 2: Unsupervised multi-modal FL; Stage 3: Weakly supervised multi-modal FL.
• Figure 5: Contrastive fusion learning on nodes with unlabeled multi-modal data. Through contrastive learning on augmented fused features, the feature encoders are trained to capture consistent information.