Automatic Diagnosis of Bipolar Disorder Using Optical Coherence Tomography Data and Artificial Intelligence

Area of Science:

• Ophthalmology
• Neuroscience
• Medical Imaging

Background:

• Bipolar disorder (BD) diagnosis currently lacks objective biomarkers.
• Optical coherence tomography (OCT) offers high-resolution retinal imaging.
• Artificial intelligence (AI) can analyze complex medical data for diagnostic support.

Purpose of the Study:

• To investigate an objective method for diagnosing BD using OCT.
• To apply AI algorithms to analyze retinal structural data in BD patients.

Main Methods:

• Retinal layer thickness was analyzed in 17 BD patients and 42 controls using OCT.
• Structural data were analyzed across ETDRS chart defined areas.
• Machine learning classifiers (Gaussian Naive Bayes, KNN, SVM) were trained on retinal features.

Main Results:

• BD patients exhibited significant retinal thinning across multiple layers compared to controls.
• Parafoveolar retinal thickness, particularly in ganglion cell and internal plexiform layers, effectively discriminated BD subjects (AUC=0.82-0.83).
• An AI classifier achieved 0.95 accuracy using inner nasal and inner inferior retinal layer thickness.

Conclusions:

• Structural retinal alterations are present in individuals with BD.
• AI analysis of OCT data shows potential as a diagnostic tool for BD.
• Larger-scale studies are required to validate these findings.