Jeadie / ENGG4801

ENGG4801 Thesis: Bayesian Deep Learning on Longitudinal Medical Data for Cancer Prognosis

ENGG4801

ENGG4801 Thesis: Bayesian Deep Learning on Longitudinal Medical Data for Cancer Prognosis

Overview

This repo contains the code used within my thesis project undertaken for my Bachelors of Engineering (Honours).

Repo Structure

The project is based on milestones of work and this repo follows such a layout. Upon progress, shared code will be taken out into shared directories. below gives the structure:

ENGG4801
│   README.md
|
└───m1_ISPY_processing/
│   │  ...
│   └── README.md
│
└───m2_ISPY_VGG/
│   │  ...
│   └── README.md
│   
└──shared/
   |  ...
   └── README.md

Where:

• m1_ISPY_processing: Is responsible for parsing the TCIA ISPY dataset from GCPs HealthCare API collection (see dataset), converting to appropriate TFRecord and distribution metadata, and egressing to AWS storage for training.
• m2_ISPY_VGG: Is responsible for reproducing the results of a paper that trained a VGG model on the ISPY1 dataset to predict NAC response (see Notion Documentation for details).
• shared: contains shared code used throughout the repo.