Computing-for-Health-and-Medicine-Final

Final Project Group 9: Diabetic Retinopathy Detection: Identifying the Severity of Diabetic Retinopathy in Eye Images
Brent Garey, Kelly Ly, Ann Men, Bishoy Sargius, and William Zouzas
COMP.4600/5300 Computing for Health and Medicine
4/26/2022

GitHub Repository:

https://github.com/lykelly19/Computing-for-Health-and-Medicine-Final

Overview:

Our goal was to create a model to efficiently and accurately detect the severity of DR in eye images to avoid manual classification by clinicians. We were able to pre-process, augment and split eye images into 4 quadrants. We created Quadrant Based Ensemble Inception V3 (instead of V2) models for each quadrant. We trained and tested the model to detect the severity of DR in eye images.

Sources:

Source used for dataset: [1] "Diabetic Retinopathy Detection." https://www.kaggle.com/competitions/diabetic-retinopathy-detection/ (accessed April 4, 2022).

Source used for implementation method using Quadrant Based Ensemble Inception: [2] C. Bhardwaj, S. Jain, and M. Sood, "Deep Learning-Based Diabetic Retinopathy Severity Grading System Employing Quadrant Ensemble Model," (in eng), J Digit Imaging, vol. 34, no. 2, pp. 440-457, Apr 2021, doi: 10.1007/s10278-021-00418-5.

Libraries used:

Matplotlib
Pandas
Pillow
NumPy
OS
Shutil
Random
Imutils
Seaborn
Statistics
Keras
Tensorflow
PyTorch
Torchvision

Required pip installs to run our code:

opencv-python
imutils

File directory:

data/
    |_labeled_data/
        |_0/
        |_1/
        |_2/
        |_3/
        |_4/
    |_normalized_whole_images/
        |_0/
        |_1/
        |_2/
        |_3/
        |_4/
    |_quadrants/
        |_quadrant_1/
            |_0/
            |_1/
            |_2/
            |_3/
            |_4/
        |_quadrant_2/
            |_0/
            |_1/
            |_2/
            |_3/
            |_4/
        |_quadrant_3/
            |_0/
            |_1/
            |_2/
            |_3/
            |_4/
        |_quadrant_4/
            |_0/
            |_1/
            |_2/
            |_3/
            |_4/

How to compile and execute our code:

Open the “DR Detection Group 9.ipynb” file using Jupyter Notebook. Click on Cell > Run All.