Federated AI for Health: Empowering Healthcare with Distributed AI Services

Slide Note

Explore how Federated AI for Health utilizes a smartphone app and trustworthy AI platform to identify symptomatic and asymptomatic patients, ensuring prompt health status alerts. Leveraging secure data collection, storage, and analysis through Machine Learning models in a federated edge computing architecture, this project aims to revolutionize healthcare access and monitoring.

adi_kle Follow

Uploaded on Sep 28, 2024 | 0 Views

Download Presentation

Please find below an Image/Link to download the presentation.

The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. Download presentation by click this link. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.

E N D

Presentation Transcript

Federated AI for Health Ana de Almeida and Maur cio Breternitz, 2023 March 31 T09:40 & 20:20 1 28-Sep-24

Motivation FCT project What is it o A smartphone app and trustworthy Artificial Intelligence distributed service-based platform For what o To identify symptomatic and asymptomatic patients as well as alert the user for a meaningful alteration of health status 2 28-Sep-24

Motivation FCT project How o Smartphone collects data, including health data, securely stores it, with access controlled and authorized by the user in his/her own device o Data is analysed by Machine Learning model in a periodic basis o Provides an AI secure backend by adopting a federated edge computing architecture based on micro-services 3 28-Sep-24

Federated AI for Health MPAI - AIH MPAI project supporting: collection of AI-based processing of access to health related data. using secure federated learning and with distribution of updated AI Models 4 28-Sep-24

MPAI approach to standardisation AI AI Open market of components with standardises functions and Interfaces, competing in performance. Module (AIM) Module (AIM) Outputs Inputs User Agent AI Workflow (AIW) AI AIM Storage AI Module (AIM) Module (AIM) Controller Global Storage MPAI Store Communication Access MPAI-AIF enables independently sourced AI Modules having standardised interfaces to be executed in an environment with standardised APIs. 5 28-Sep-24

MPAI-AIH main Actors End-User Third-Party External Data Source MPAI-AIH Services Front-End System Back-End System Blockchain & Distributed Ledgers AI Services MPAI-AIH Intelligent Computational Service Organization Centralized Federated Learning Federated Learning Clients: Mobile Applications Multisourced AIs (AIMs) 6 28-Sep-24

Access to multiple services from smartphone (data storage, permissions, licenses, etc.) MPAI-AIH Architecture Blockchain and Distributed Ledger Technologies Licensing & Governance Services Authentication & Access Control Services Record of transactions on the B&DLT that can be used for audition purposes Auditing Services API API Other sources of health- related data De-Identification and Anonymization Services Data Storage & Access Services Global Secure Data Vault Third-Party Health-related Entities AI Module (AIM) AI Module (AIM) Access to data (de- identified & anonymized) Processing of data through AI to extract knowledge User Agent AI Workflow (AIW) AIM Storage AI Module (AIM) AI Module (AIM) Controller MPAI Store Communication Global Storage Access Encryption Service AIM Storage Service Communication Service Attestation Service AIM Model Service AIM Security Engine Trusted Services 7 28-Sep-24

Anomaly Detection & Risks Alert End-user input data Local AIH FL Model Specific Biometric Sensors Biometric Signals Smartbands, Smartwatches, Other health- related apps installed Interaction with AIH System Backend and Centralized FL Model Services AI-Health App Secure Data Vault The Front-End System 8 28-Sep-24

Federated Learning using Mobile applications Federated Learning is a technique that enables ML algorithms deployed across multiple decentralized edge devices or servers holding local data samples to collaboratively train a global model. No user data exchange, only (incremental) changes from the local machine learning model are uploaded to the global machine learning model. Eventually, upon authorization, a new (updated) model is downloaded into the edge devices, in an exchange between the decentralized devices and a server. 9 28-Sep-24

Federated Learning training process 10 28-Sep-24

Functional requirements 1. An AIH Platform Front-end shall be persistent, i.e., the Front End runs continuously. 2. AIH Platform Front-end biometric sensor calibration 3. Terms and conditions in a Smart Contract: Front-End to Back-End & Third Part to Back- End 4. Processing requested by Third Parties 5. Data types possible target of processing 6. Interface between: a. Auditing and Licensing & Governance Services b. Authentication & Access Control Services and Auditing Services c. Authentication & Access Control Services and Licensing & Governance Services d. Interface Authentication & Access Control Services and Data Storage & Access Services e. Data Storage & Access Services and De-Identification & Anonymization Services f. Back-End and the AIF 11 28-Sep-24

Where are we in the MPAI process Stage 0 Stage 1 Stage 2 Stage 3 Interest Collection Use Cases Functional Requirements Commercial Requirements Technical Specification Reference Software Call for Technologies Community Comments Standard Development Conformance Testing Reference Software Performance Assessment Stage 7 Stage 5 Stage 4 Stage 6 Principal Members All All Members 12 28-Sep-24