1. Rhino Health
  2. Learning How to Use FCP with Rhino Tutorials
  3. Sandbox
  4. Pneumonia Prediction

Pneumonia Prediction Step 1 - Scenario & FCP Overview

Updated

 

Rhino Health Sandbox.png

Welcome to the Rhino Health FCP Sandbox!

Pneumonia Prediction

 

We're so excited you're here. Enjoy this demonstration!

 

Here you will learn to leverage the FCP to access federated datasets and work on federated data processing, metrics, and AI modeling and evaluation on one trusted computing platform.

sandbox_overview.png

Above is a diagram (click on the image to enlarge it) outlining the next 5 actions you will perform as part of this 6-step pneumonia prediction use-case.  Each step is completed by following along with the documentation here and working within the associated Jupyter notebooks.  The code artifacts for this use case can be found here. See you at the finish line!

 

The Scenario

Dr Emily Smith.png

Meet Dr. Emily Smith, a hospital physician with a passion for harnessing technology to improve patient outcomes. Today, we'll join her as she embarks on a mission to build a life-saving pneumonia prediction model.

The Technology

Dr. Smith, interested in exploring Federated Learning (FL), immediately gravitated towards a federated architecture for her project, explicitly citing some of the numerous benefits of FL:

  • Collaboration across different organizations and data silos enabling her to train her pneumonia model with more diverse data
  • Scalability that would allow collaborators to quickly onboard and contribute to her project, aiming for more accurate modeling
  • Privacy preservation by keeping data local within institutional firewalls with no need for centralized storage
  • Enhanced patient privacy no raw data is ever transmitted; only model weights/parameters and k-anonymized data aggregates

The Platform

She knew exactly which platform to best execute her project on—the Rhino Health Federated Computing Platform (FCP). The Rhino Health FCP lets AI developers work on decentralized data as if it were centralized by offering an extensible, privacy-preserving platform that can facilitate full end-to-end AI workflows, including exploratory analysis, pre-processing and harmonization, annotation, model training, and deployment of custom code, among others.
Dr. Smith realized she needed to understand the various core components within the FCP to be successful in her project.

  1. The Project - a foundational unit for secure, collaborative computing and federated learning initiatives on the FCP. It encompasses Datasets, Data Schemas, Code, and Collaborators, offering a structured framework for impactful research and innovation, all anchored by unwavering data privacy and security.
  2. The Data Schema - a blueprint for data, defining structure, attributes, and organization for consistent, standardized analysis
  3. The Dataset - a secure dataset linked to a specific project, owned by a particular workgroup, and stored on their Rhino Client. It holds various data formats (tabular, file, DICOM) and is organized by a defined Data Schema.
  4. The Code - the basic computational building block in a project, ranging from a simple pre-processing script to an intricate federated learning pipeline. It allows you to transform data, train models across distributed sites, and even run specialized 3rd-party software, all within the platform's secure environment.
  5. The Code Run - a hub for all vital information about a specific Code run, offering a complete picture of its execution, logs, and outcomes.
  6. The Collaborator - members of a workgroup who are granted controlled access for streamlined collaboration within a specific project by a project owner.
  7. The Rhino SDK - the Python toolkit for unlocking the platform's power within your own development environment.  Containing equivalent functionality to the FCP UI, users can accomplish their task directly through Python code.

Getting Started

Once Dr. Smith had a good understanding of the core components within the FCP, her next step would be to prepare her local development environment before beginning to work on her pneumonia project. That includes steps like installing dependencies (Docker, AWS Command Line Interface (CLI), and Amazon Elastic Container Registry (ECR) Credential Helper), retrieving SFTP and ECR credentials, and configuring the AWS CLI.

Additional Resources

She noted several other official resources that might come in handy as she progresses with her project:

 

If you need support at any time, feel free to contact support@rhinohealth.com.

 

Continue to Pneumonia Prediction Step 2 →

Was this article helpful?
0 out of 0 found this helpful

Articles in this section