F23_HealthChain

Project Description

HealthChain aims to create a framework for standardizing personal health data using blockchain technologies and allow for long-term health data analytics through generative AI to allow for objective referral information.

Instructions

• Clone the repository onto your local machine
• Open a terminal/command prompt and navigate to the UI directory of the repository
• pip install the requirements using 'pip install -r requirements.txt'
• Run the following command 'python manage.py runserver'
• Navigate to http://127.0.0.1:8000/ using an internet browser

AI component

• Data: Synthea is a free open-source software that allows for the generation of synthetic patient medical records that are reflective of actual records, allowing training on large synthetic datasets
• Considering the sparse time-series nature of medical records (visits can be rare, inconsistent, and with variable observations), a sequential model with an attention mechanism like a transformer seems like a logical choice to navigate the inconsistent data: important observations can be recognized at different times for different patients
  • Due to sparsity, time-steps may have to be aggregates of visits, and/or patients with limited data may need to be excluded
  • Adding engineered/derived features could be useful e.g. age from visit date and birthdate, or rate of observation increase/decrease over time
• AI trains on synthetic patients with binary cross-entropy for classification so that it can predict likelihoods of diseases in patients
• Disease probabilities are passed to an algorithm for querying a database of doctors: specializations, ratings etc. would be utilized to determine referrals

Blockchain component

• Permissioned Blockchain: The blockchain will be a permissioned blockchain. The medical records need to be associated with the patients and doctors need to be verified to ensure the safety of the patients.
• Oracle:
  • We will need an oracle to fetch the medical records from an off-chain database as we do not want the medical records to be stored on the blockchain for fear of privacy issues.
  • We’ll also need another Oracle to communicate with the AI models
• The blockchain would serve as a governing factor to control permissions to edit and view patient data
• Patients would be able to allow doctors to edit personal data.
• Patients would also be able to remove doctor permission once they saw fit
• The blockchain would consist of a series of smart contracts to allow this functionality to persist as well as supporting backend capabilities through the usage of maps primarily

High-Level Component Diagram

Sequence Diagram

sequenceDiagram
    participant A as Patient
    participant B as UI
    participant C as Smart Contract
    participant Z as Doctor
    participant E1 as Oracle for IPFS
    participant D as IPFS
    participant E as Oracle for AI
    participant F as TNN for Disease Identification
    participant G as Algorithm for Doctor Referral
    participant H as Doctor Database

    A->>B: Add/Remove Doctors or View MR or AI Request
    Z->>B: View MR or Update MR
    B->>C: Forward Request
    C->>E1: Retrieve or Update MR
    E1->>D: Access IPFS
    D->>E1: Return MR and/or New Hash
    E1->>C: Send MR and/or Update Hash
    C->>B: Decrypt MR for View
    B->>A: Decrypted MR for View
    B->>Z: Decrypted MR for View/Update
    C->>E: AI Request w/ MR
    E->>F: Forward AI Request
    F->>G: Disease Probabilities
    G->>H: Query
    H->>G: Identify Specialist(s)
    G->>E: Report
    E->>C: Report
    C->>B: Report
    B->>A: Referall(s)

Loading