NA-MIC Project Weeks

Back to Projects List

Risk Prediction Deep Learning Models into MHub.ai

Key Investigators

• Ahmed Adly (Harvard Medical School, USA)
• Leonard Nuernberg (Harvard Medical School, USA)
• Hugo Aerts (Harvard Medical School, USA)

Presenter location: In-person

Project Description

Mhub.ai is a framework to enhance reproducible research by standardizing models into Mhub containers that could be flexible and effortless to use. Therefore I aim to add two often used risk prediction models (Sybil and CVD-risk-estimator), to make it easy for the community to run such models through Mhub using a standardized way (by one simple command).

Objective

1. Getting more familiar with Mhub.ai framework, to keep pushing high quality models there for reproducible science.
2. Publishing risk prediction Models on Mhub.ai .

Approach and Plan

1. Attending MHub workshop held at PW, so that I grasp best practices.
2. Start with a basic hands on -> Mhub.ai converter from DICOM to NRRD.
3. Wrap the risk prediction models (Sybil / CVD-risk-estimator) for Mhub.ai Framework.
4. Run the models on data using Mhub.ai and Github, to compare the simplicity of the approach, efficiency (time and effort) and output.

Progress and Next Steps

Before PW

1. Getting more familiar with Mhub.ai infrastructure and documentation.
2. Going through Mhub.ai tutorials.

After PW

1. Sybil - Cancer risk prediction Model - is wrapped in MHUB.ai Framerwork and pushed to Mhub.ai.
2. CVD-Risk-Estimator - CVD risk model - is still ongoing, however at last stage.
3. Got More comfortable with Mhub.ai framework, and looking forward to add more models.

Illustrations

No response

Background and References

• Mhub
• Sybil-Cancer Prediction Model
• CVD-Risk-Estimator