Overview

In this assignment, you will explore a) classification, b) segmentation, and c) 3d segmentation and semi-supervision on medical images. In part a, you will predict severity of diabetic retinopathy in patients using retina images. In part b, you will explore segmenting lungs in computerized tomography (CT) scans. In part c, you will modify your lung segmentation model to achieve semi-supervised 3D lung segmentation.

Assignment Setup Instructions

We will need to increase GCP disk space to fit all the data for this assignment. To do so for a VM instance, please do the following:

1. Go to Compute Engine > Disks.
2. Click into your VM instance and click EDIT.
3. In the Size field, type in "200" GB.
4. Click SAVE at the bottom of the page.

Remember to always stop your VM instance when you're not using them, by following the instructions in the assignment 0 GCP Guide.

It has come to our attention that sometimes the GPU in a GPU instance may not be recognized (e.g. nvidia-smi not working). If this is the case, we recommend using a GPU notebook instance (see Section 5 of the Google Cloud guide). We recommend a notebook server with at least 8 CPUs, 30GB memory, and 300 GB disk space.

Download the assignment files on your VM instance and then unzip them:

                wget https://biods220.stanford.edu/notebooks/assign1.zip
                unzip assign1.zip
              

Once you unzip, you will notice that there are three notebooks located in assign1/. Each notebook is for one part of the assignment - part a) diabetic retinopathy, b) 2D lung segmentation, and c) 3D lung segmentation. Complete both the coding and written sections. To do this, you will need to launch Jupter Notebook as stated in these instructions (from assignment 0). Good luck on Assignment 1!

What to hand in for this assignment: Submit one PDF file containing all your notebook solutions (code + written). To convert your notebook to PDF, go to File > Download as > PDF via LaTeX. Alternatively, if you encounter issues with converting to PDF using LaTeX, you can also convert to HTML (File > Download as > HTML) and then "Save to PDF" by printing. Concatenate the PDFs from all three parts into one and make sure that both the coding sections and the written sections are visible in the PDFs prior to submission. Please make sure when submitting to Gradescope that all questions have been marked with the appropriate pages.