Comparison between support vector machine and deep learning, machine-learning technologies for detecting epiretinal...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Epiretinal membrane (ERM) is a common condition affecting vision.
• Accurate detection of ERM is crucial for timely diagnosis and treatment.
• Three-dimensional optical coherence tomography (3D-OCT) provides detailed cross-sectional retinal images.

Purpose of the Study:

• To compare the efficacy of deep learning (DL) and support vector machine (SVM) algorithms in detecting ERM.
• To evaluate the performance of DL and SVM using 3D-OCT imaging data.

Main Methods:

• A dataset of 529 3D-OCT images (184 non-ERM, 205 ERM) was utilized.
• Images were split into training (80%) and testing (20%) sets.
• DL (deep convolutional neural network) and SVM models were trained and evaluated.

Main Results:

• The DL model achieved a sensitivity of 97.6% and specificity of 98.0% with an AUC of 0.993.
• The SVM model achieved a sensitivity of 97.6% and specificity of 94.2% with an AUC of 0.988.
• DL demonstrated higher specificity and AUC compared to SVM.

Conclusions:

• Deep learning models show improved accuracy in detecting ERM compared to SVM.
• DL offers a promising tool for automated ERM detection using 3D-OCT.