1. Functions | 2. Reference Model | 3. Input/Output Data |
4. Functions of AI Modules | 5. Input/output Data of AI Modules | 6. AIW, AIMs, and JSON Metadata |
1. Functions
The Back End gathers Health Data from various sources (End Users and External Sources) and acts as both a brokerage gateway between the entities requesting access to the Health Data and those providing it, as well as data and AI services provider. The Back End grants Rights without referring to the identity of the End Users who have provided the Health Data as long as Health Data is licensed by Smart Contracts registered on the Block Chain.
The Back End implements the following functionalities:
- Securely store, de-identify and anonymise Health Data.
- Authentication and access control to Health Data.
- Receive and store licensed Health Data.
- License access to Health Data.
- Receive requests and process Health Data through the AI Framework.
- Perform Federated Learning services to update Font-Ends.
- Upload and download AIMs from the MPAI Store.
- Communicates with the Blockchain.
2. Reference Model
Figure 1 depicts the Reference Architecture of the Health Back End (AIH-HBE).
Figure 1 – Reference Model of the Health Back End (AIH-GBE) AIW
This is the flow of operation of the Health Back End (AIH-GBE) AIW Reference Model:
- A User wishing to access the Back End, provides Authentication and Access Control Parametres to Authentication and Access Control (MPAI-AAC) AIM.
- If Authentication is successful, MPAI-AAC AIM issues a Token. The following steps of Health Back End access assume that a User holds a Token.
- User uploads Health Data with Model Licence.
- Health Data is received by the Data Storage and Access (AIH-DSA) and stored to the Secure Storage.
- The Model Licence is converted to a JSON format and stored in the Block Chain as a Smart Contract.
- A User may make a De-Identification and Anonymisation Request to the De-Identification and Anonymisation (AIH-DIA) AIM:
- AIH-DIA AIM retrieves Health Data from Secure Storage.
- De-identifies and anonymises Health Data.
- Stores DI&A’d Health Data back to Secure Storage.
- A User may make a Health Data Processing Request to the Health Data Processing (AIH-HDP) AIM:
- AIF Controller may download appropriate AIM from the MPAI Store.
- AIH-HDP
- Request an appropriate Licence from the Block Chain.
- Retrieves Health Data from Secure Storage.
- Performs requested processing.
- Stores processed Health Data to Secure Storage.
- Outputs processed Health Data.
- A User may make an Audit Request to the Audit (AIH-ADT) AIM providing Audit Data, the AIH-ADT AIM
- Requests AIH Data Statistics and Queries to the Block Chain.
- Provides Audit Response.
- When Federated Learning process begins:
- Health Back End requests the NN Model to a Health Front Ends.
- Health Front Ends provide the NN Models.
- Health Back End
- develops and upload to the MPAI Store the new NN Model.
- Informs the Health Front Ends.
- Front Ends download the new NN Model from the MPAI Store.
3. Input/Output Data
Table 2 specifies the Input and Output Data.
Table 1 – I/O Data of Health Front End
Input | Description |
AAC Parameters | Parameters provided by User to the Authentication & Access Control AIM. |
AIH Data Access Request | End User’s request to Upload or Access AIH Data. |
AIH Data | Health Data provided by User’s Front End. |
Model Licence | Data describing the licence terms and conditions. |
DI&A Request | Request to de-identify and anonymised AIH Data. |
Processing Request | Request to process AIH Data. |
Audit Request | Audit Request made by End User. |
Audit Data | Data required bt the Audit AIM. |
NN Module | NN Model provided by Front End. |
Output | Description |
AAC Token | Token provided to User by Authentication & Access Control AIM is process was successful. |
Model Licence | Licence represented according to Licence JSON Schema. |
DI&A Data | Data that has been de-identified and nonymised. |
Licence Request | Request made ro Block Chain to enable AIH Data Processing. |
Processed Data | AIH Data that have been processes. |
Stats & Query Request | Request made to Block Chain to perfor Aidit. |
Audit Response | Response containing Audit results. |
NN Module | NN Model provided byFederated Learning. |
NN Module Request | Request made to Front End by Back End for Federated Learning. |
4. Functions of AI Modules
Table 2 specifies the Function of the AI Modules.
Table 2 – Functions of Health Front End AI Modules
AI Module | Description |
Authentication & Access Control | Produces Token to User who have successfully passed the process. |
Data Storage and Access | Manages AIH Data upload and subsequent Access. |
De-Identification & Anonymisation | De-identifies and nonymises AIH Data. |
AIH Data Processing | Processes AIH Data |
Auditing | Emables End User to monitor the use of their AIH Data. |
Health Federated Learning | Performs Federated Learning of Front Ends’ NN Model. |
5. Input/output Data of AI Modules
Table 2 specifies the Function of the AI Modules.
Table 3 – Functions of Health Front End AI Modules
AI Module | Receives | Produces |
Authentication & Access Control | AAC Parameters | AAC Token |
Data Storage and Access | AIH Data Access Request AIH Data |
AIH Data |
De-Identification & Anonymisation | DI&A Request AIH Data |
DI&A Data |
AIH Data Processing | Processing Request AIH Data |
Processed Data |
Audit | Audit Request Audit Data |
Stats & Query Req. Audit Response |
Health Federated Learning | NN Module | NN Module NN Module Request |
6. AIW, AIMs, and JSON Metadata
Table 4 provides the links to the AIW and AIM specifications and to the JSON syntaxes. AIMs/1 indicates that the column contains Composite AIMs and AIMs indicates that the column contains their Basic AIMs.
Table 4 – AIW, AIMs, and JSON Metadata
AIW | AIMs | Name | JSON |
AIH-HBE | Health Back End | X | |
AIH-AAC | Authentication & Access Control | X | |
AIH-DSA | Data Storage and Access | X | |
AIH-DIA | De-Identification & Anonymisation | X | |
AIH-HDP | AIH Data Processing | X | |
AIH-ADT | Audit | X | |
AIH-HFL | Health Federated Learning | X |