Acne Lab

A MATLAB-based intelligent application that automates the global acne grading assessment.

Comment

What it does

This project makes use of MATLAB deep network designer to create and train two deep learning models. They classify the type of acne and the anatomical sites from images. Based on the global acne grading system [1], the MATLAB app interface calculates the local score for each affected area and displays the global score. The global score is the sum of local scores, and the acne severity was graded using the global score.

Inspiration

  • Acne is very common among adolescents and young adults across all cultures. Without proper assessment and care, the problem can persist into adulthood.
  • Acne is one of the most common dermatologic conditions. Over 85% of teenagers are affected by acne at some point during their teenage years.
  • Most of the time, acne affects the face, the upper part of the chest and the back.

Acne <--> Negative impact on self-image <--> Mental health issues

"Acne has been associated with higher rates of depression, anxiety, failure to thrive at school and in social environments, suicidal ideation, and suicidal attempts" (Halvorsen et al., 2011, Picardi et al., 2006, Purvis et al., 2006, Yentzer et al., 2010).

Bridge the gap between those suffering from acne and proper medical treatment.

This application can...

  1. Reduce the cost of visiting a dermatologist
  2. Minimize inter and intra-observer variability between different dermatologists in characterizing acne.

How I built it

This project (the machine learning model and the app interface) has been built purely with MATLAB.

Accomplishments that I am proud of

I experienced a lot of first-time in the hackathon

  1. First time working on a machine learning project
  2. First time using MATLAB app designer
  3. First time training a deep network
  4. First time hacking alone

What's next for Acne Lab

  • Tuning the initial learning rate, epoch and batch size to achieve higher accuracy.
  • Collect more images of acne.
  • Set up a system to provide treatment, life style, food, or skin care routine recommendations.

How to run my project code

Prequisites

The nature of taking images of the patient's face in the app interface requires "MATLAB Support Package for USB Webcams" link

Try out procedure

  1. Download trainedNetwork_1.mat. It contains everything needed to run the GUI.
  2. Download MATLAB Support Package for USB Webcams if you haven't.
  3. Run Acne_Lab2.mlapp in you local MATLAB. Make sure .mat file is in the same folder.
  4. Click turn on camera or upload image.
  5. If you turn on the camera, position yourself so that the camera is capturing your acne. Click take a picture once you are ready.
  6. The computer will run the two classification models and generate the results on the GUI.
  7. Click update total score
  8. Repeat step 3 to 6 if you have more than one acne-affected area

[1] Doshi, A., Zaheer, A. and Stiller, M.J. (1997), A comparison of current acne grading systems and proposal of a novel system. International Journal of Dermatology, 36: 416-418. https://doi-org.myaccess.library.utoronto.ca/10.1046/j.1365-4362.1997.00099.x

Built With

Share this project:

Updates