Related Experiment Video
Updated: Jun 12, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Computer Vision Identification of Trachomatous Inflammation-Follicular Using Deep Learning
Ashlin S Joye1,2, Marissa G Firlie3, Dionna M Wittberg2
1Casey Eye Institute, Oregon Health and Science University, Portland, OR.
Cornea
|September 23, 2024
Summary
Machine learning models accurately detect trachoma inflammation-follicular (TF) from eye images, offering a reliable and cost-effective alternative to human grading for global health surveys.
Area of Science:
- Ophthalmology
- Public Health
- Artificial Intelligence
Background:
- Trachoma surveys are crucial for estimating disease prevalence and guiding antibiotic distribution.
- Current methods rely on human graders, which are resource-intensive and prone to errors.
Purpose of the Study:
- To develop and evaluate machine learning models for automated trachoma grading.
- To reduce costs and improve the reliability of trachoma screening surveys.
Main Methods:
- A deep convolutional neural network (MobileNetV3 large) was trained using 56,725 everted eyelid photographs from Ethiopian children aged 0-9 years.
- Ground truth was established using the median estimates from three expert grader groups.
Main Results:
- The model achieved high performance with an area under the receiver operating characteristic curve of 0.943, an F1 score of 0.923, 88% accuracy, 83% sensitivity, and 91% specificity.
- Predicted TF prevalence (32%) closely matched the human consensus estimate (30%).
Conclusions:
- Deep convolutional neural network models demonstrate strong performance in classifying trachoma inflammation-follicular (TF) and follicle counts from conjunctival images.
- These models show potential for accurate, efficient, and large-scale trachoma screening.
- Further validation in diverse populations is recommended before widespread implementation.
