Discrimination of multiple sclerosis using OCT images from two different centers

Area of Science:

• Ophthalmology
• Neurology
• Artificial Intelligence

Background:

• Multiple sclerosis (MS) is a chronic inflammatory disease affecting the central nervous system.
• Optical coherence tomography (OCT) is a noninvasive imaging technique with potential as a biomarker for MS.
• Subtle changes in retinal layer thickness in MS necessitate advanced analysis methods beyond raw OCT data.

Purpose of the Study:

• To develop and validate an AI-based algorithm for discriminating between MS patients and healthy controls (HCs) using multilayer segmented OCTs.
• To identify key retinal layers and topological features most indicative of MS.
• To ensure the trustworthiness and robustness of the AI model through interpretability and external validation.

Main Methods:

• Multilayer segmented OCTs were processed using dimension reduction to select discriminative features.
• Classifiers including Support Vector Machine (SVM), random forest (RF), and artificial neural network (ANN) were employed.
• Patient-wise cross-validation and occlusion sensitivity were used for performance evaluation and interpretability.

Main Results:

• The optimal feature topology was a square of 40 pixels.
• The ganglion cell and inner plexiform layer (GCIPL) and inner nuclear layer (INL) were the most influential retinal layers.
• Linear SVM achieved 88% accuracy, 78% precision, and 63% recall in distinguishing MS from HCs.

Conclusions:

• The AI classification algorithm shows promise for assisting neurologists in the early diagnosis of MS.
• The use of two distinct datasets enhances the robustness and external validity of the findings.
• The study demonstrates effective outcomes without deep learning, suitable for limited data scenarios.