System and method to diagnose conjunctivitis in the eye of a user

Area of Science:

• Medical Informatics
• Computer Vision
• Machine Learning

Background:

• Conjunctivitis is a prevalent eye condition requiring efficient diagnostic methods.
• Current diagnostic approaches may benefit from advancements in artificial intelligence.
• Automated systems can potentially improve the speed and accuracy of eye disease detection.

Purpose of the Study:

• To investigate the efficacy of machine learning models for conjunctivitis diagnosis.
• To compare the performance of UNet++ and U-net segmentation models in eye image analysis.
• To develop an automated system for conjunctivitis detection using TensorFlow.

Main Methods:

• Image acquisition using camera-based systems.
• Application of image pre-processing and segmentation techniques (UNet++, U-net).
• Feature extraction and classification using Convolutional Neural Networks (CNNs) within TensorFlow.
• Model evaluation using the UBIRIS dataset and a custom dataset.

Main Results:

• The UNet++ model demonstrated superior performance in image segmentation for conjunctivitis diagnosis.
• UNet++ achieved an overall accuracy of 97.07%.
• Comparative analysis showed UNet++ outperformed the traditional U-net model in accuracy and performance.

Conclusions:

• UNet++ represents a significant advancement for machine learning-based conjunctivitis diagnosis.
• The developed methodology shows high potential for clinical application in ophthalmology.
• Further research can explore broader applications of UNet++ in diagnosing other eye conditions.