Predicting Axial Length From Choroidal Thickness on Optical Coherence Tomography Images With Machine Learning Based...

Area of Science:

• Ophthalmology
• Medical Imaging
• Machine Learning

Background:

• Accurate measurement of axial length (AXL) is crucial for diagnosing and managing various ocular conditions.
• Optical coherence tomography (OCT) provides high-resolution cross-sectional images of the retina and choroid.
• Choroidal thickness (CT) is a potential indicator of ocular health and disease, often correlated with AXL.

Purpose of the Study:

• To develop and evaluate ensemble learning models for classifying axial length (AXL) based on choroidal thickness (CT) measurements from 2D OCT images.
• To assess the performance of these models in both binary and multiclass AXL classifications.

Main Methods:

• A retrospective cross-sectional study analyzed 710 OCT images from 188 patients.
• Ensemble learning models were constructed using five base machine-learning algorithms.
• Models were trained and validated for binary (AXL < or > 26 mm) and multiclass (AXL < 22 mm, 22-26 mm, > 26 mm) classifications.

Main Results:

• Feature selection confirmed no redundant variables, with nasal CT showing the highest AXL correlation.
• Binary classification achieved high accuracy (94.37%), recall (100%), and AUC (95.61%).
• Multiclass classification demonstrated strong performance with accuracy (88.73%) and macro AUC (93.42%).

Conclusions:

• Ensemble learning models effectively classify axial length from choroidal thickness using OCT images.
• The developed classifiers show high accuracy in both binary and multiclass AXL categorization.
• These models serve as a promising assistive tool for ophthalmologists.