High Intraocular Pressure Detection from Frontal Eye Images: A Machine Learning Based Approach

Area of Science:

• Ophthalmology
• Medical Imaging
• Machine Learning

Background:

• Intraocular pressure (IOP) is a critical metric for diagnosing and managing glaucoma.
• Current IOP measurement methods can be invasive or require specialized equipment.
• There is a need for non-invasive, accessible methods for IOP assessment.

Purpose of the Study:

• To develop and validate a novel, automated framework for detecting normal versus high intraocular pressure (IOP) using only frontal eye images.
• To explore the efficacy of machine learning in analyzing ocular features for IOP status detection.

Main Methods:

• A machine learning framework was developed to extract six features from frontal eye images.
• Features included Pupil/Iris ratio, red area percentage, scleral redness, and novel scleral contour features (angle, area, distance).
• Two classifiers, decision tree and support vector machine, were applied to a database of 400 annotated frontal eye images.

Main Results:

• The proposed framework achieved an overall accuracy of 95.5% in detecting IOP status.
• The decision tree classifier demonstrated superior performance in this task.
• The automated feature extraction and classification provide a robust method for IOP assessment.

Conclusions:

• Solely analyzing frontal eye images with machine learning can accurately detect intraocular pressure status.
• The developed framework offers a promising, non-invasive, and automated approach for eye pressure monitoring.
• This technology has the potential to improve early detection and management of conditions related to elevated IOP.