Diagnosis of multiple sclerosis using optical coherence tomography supported by artificial intelligence

Area of Science:

• Ophthalmology
• Neurology
• Biomedical Engineering

Background:

• Current multiple sclerosis (MS) diagnostic methods have limitations.
• Identifying novel biomarkers is crucial for early MS detection.
• Spectral-domain optical coherence tomography (OCT) offers potential for biomarker discovery.

Purpose of the Study:

• To identify new biomarkers for the early diagnosis of MS.
• To utilize spectral-domain OCT and artificial intelligence for MS diagnosis.
• To evaluate the diagnostic capability of neuroretinal structure analysis.

Main Methods:

• Spectral-domain OCT was performed on 79 relapsing-remitting MS patients and 69 healthy controls.
• Retinal thickness and inter-eye differences were analyzed in various retinal layers.
• A convolutional neural network was employed to assess diagnostic capability using OCT data.

Main Results:

• Significant alterations in retinal thickness were observed in the ganglion cell (GCL), inner plexiform (IPL), and inner retinal (IRL) layers of MS patients.
• The average GCL thickness (AUROC = 0.82) and IPL inter-eye difference (AUROC = 0.71) were key discriminant features.
• An artificial intelligence model achieved an accuracy of 0.87, sensitivity of 0.82, and specificity of 0.92 for MS diagnosis.

Conclusions:

• Neuroretinal structure analysis shows promise as a diagnostic tool for MS.
• OCT-based biomarkers could potentially be integrated into future MS diagnostic criteria.
• This study highlights the role of AI in analyzing OCT data for neurological disease diagnosis.