Artificial Intelligence in Fluorescence Lifetime Imaging Ophthalmoscopy (FLIO) Data Analysis-Toward Retinal Metabolic...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Fluorescence lifetime imaging ophthalmoscopy (FLIO) provides functional information about the retina.
• Analyzing FLIO data, especially from healthy subjects with small datasets, presents a challenge.
• Artificial intelligence (AI) offers potential solutions for complex data analysis in medical imaging.

Purpose of the Study:

• To investigate the feasibility of using AI for analyzing FLIO data, particularly with small sample sizes.
• To evaluate the effectiveness of different AI methods, including support vector machines (SVMs), convolutional neural networks (CNNs), and autoencoder networks.
• To assess AI's capability in differentiating between smokers and non-smokers based on FLIO parameters.

Main Methods:

• Utilized FLIO data (fluorescence intensity, mean fluorescence lifetime τm) and OCT-A data from 26 non-smokers and 28 smokers.
• Applied SVM, CNN, and autoencoder networks for data analysis.
• Focused on analyzing τm in two spectral channels.

Main Results:

• SVM successfully distinguished mean fluorescence lifetime (τm) between non-smokers and heavy smokers with approximately 80% accuracy.
• CNN and autoencoder networks did not achieve significant differentiation.
• Optical Coherence Tomography Angiography (OCT-A) data showed no significant differences between groups.

Conclusions:

• AI, particularly SVM, is feasible and useful for analyzing FLIO data from healthy subjects, even with small datasets.
• AI-assisted FLIO shows promise for advancing early retinal diagnosis.
• Further validation with larger datasets is recommended to confirm these findings.