Related Experiment Video
Updated: Jul 16, 2026
05:51
Smartphone Fundus Photography
Published on: July 6, 2017
Smartphone-Assisted Placido Ring Imaging for K1 Stratification in Keratoconus: A Deep Learning Study
Enes Eroglu1, Nicholas Tomaras2, Kabir Anand Pathak1
1Department of Ophthalmology, University of North Carolina at Chapel Hill, 2226 Nelson Highway, Chapel Hill, NC 27517, USA.
Diagnostics (Basel, Switzerland)
|July 15, 2026
Summary
A new deep-learning algorithm uses smartphone images to categorize keratoconus severity. This low-cost method shows high accuracy in predicting keratometric categories, potentially improving accessibility for diagnosing this eye disease.
Area of Science:
- Ophthalmology and computational vision
- Artificial intelligence in medical diagnostics
Background:
- Keratoconus (KC) is a progressive corneal disease impacting vision.
- Standard diagnostic tools like corneal topography are not always accessible.
- Need for low-cost, accessible KC diagnostic methods.
Purpose of the Study:
- To evaluate a novel, low-cost deep-learning algorithm for inferring keratometric categories from smartphone-assisted Placido ring photographs.
- To assess the feasibility of using smartphone imaging for KC stratification.
Main Methods:
- A Variational Autoencoder (AutoEncoderKL) was trained on healthy and KC eye images.
- An ensemble classifier utilized encoder features to predict K1 keratometric categories (<40 D, 40-47 D, >47 D).
- Performance evaluated using accuracy, precision, recall, and F1-score on a held-out set.
Main Results:
- The model achieved 91% overall accuracy across all keratometric categories.
- High precision and recall were observed, with perfect recall for the >47 D category.
- The algorithm successfully predicted K1 keratometric categories without direct tomographic or keratometric inputs.
Conclusions:
- Smartphone-assisted Placido ring imaging combined with deep learning offers a promising low-cost approach for KC stratification.
- This method provides a proof-of-concept for accessible KC screening.
- Further validation in diverse clinical settings is necessary to assess full clinical utility.
