Classification algorithms based on anterior segment optical coherence tomography measurements for detection of angle...

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• Anterior segment optical coherence tomography (ASOCT) parameters can explain over 80% of variability in angle width.
• Gonioscopic angle closure is a significant clinical finding requiring accurate detection methods.

Purpose of the Study:

• To evaluate the effectiveness of classification algorithms using ASOCT measurements for detecting gonioscopic angle closure.
• To identify the optimal algorithm for classifying angle closure based on ASOCT data.

Main Methods:

• A cross-sectional study involving 2047 subjects aged 50 years and older.
• Six classification algorithms were tested using ASOCT-derived parameters, including stepwise logistic regression, Random Forest, and support vector machine.
• 10-fold cross-validation and 50:50 training/validation methods were employed to assess algorithm performance.

Main Results:

• Stepwise logistic regression demonstrated the highest accuracy in detecting gonioscopic angle closure, achieving an area under the receiver operating characteristic (AUC) curve of 0.954 and 0.962 in different validation sets.
• The best-performing algorithm correctly identified subjects with gonioscopic angle closure over 95% of the time.
• Recursive partitioning showed the poorest performance, with AUCs ranging from 0.860 to 0.905.

Conclusions:

• A classification algorithm based on stepwise logistic regression, utilizing 6 ASOCT parameters from a single scan, effectively identifies gonioscopic angle closure.
• This ASOCT-based approach offers a highly accurate, non-invasive method for diagnosing angle closure, with potential clinical applications.