Group Research Project

Lupus Detection System

Start date: 2024-08-29

Written by: Abdilahi

MNSU Systemic Lupus Erythematosus Detection System

Project Description:

Systemic Lupus Erythematosus (SLE) is a chronic autoimmune disease that affects millions worldwide, particularly women aged 15-44.One key indicator of SLE is the Butterfly Malar Rash (BMR), a distinctive rash across the cheekbone and nasal bridge.

This project aimed to utilize convolutional neural network (CNN) architectures (e.g. ResNet, VGG) in designing early SLE detection systems. The project used a dataset of facial images with and without BMR to train and test the CNN models. The project explored the use of transfer learning to improve model performance.

Objective:

1. Create a dataset containing facial images of BMR and other similar rashes. 
2. To construct the dataset an Extract-Transform-Load (ETL) pipeline will be designed, 
   that will fetch image data from online sources.
3. Perform filtering and preprocessing and then store them based on their respective classes.
4. Development of a state-of-the-art CNN based model to detect BMR from facial images.
5. Utilize Transfer learning will be in designing CNN based model for BMR detection. e.g ResNet, VGG, and  DenseNet.
6. Draft a conference and/or journal manuscript for publication.

Sample Dataset Image

Dataset Source Distribution

Attempted Machine Learning Models

Below are the pretrained models and performance metrics on our dataset.

Research Paper

Our research paper got accepted for the IEEE Conference CAI 2025 - Vertical AI for Healthcare and Life Sciences.

Awards

My team and I came first in the CADSCOM 2024 Project Poster.