AI-enhanced 3D-OCT analysis for automated mapping of layer thickness in paintings

Area of Science:

• Art diagnostics
• Non-invasive imaging techniques
• Cultural heritage research

Background:

• Characterizing heterogeneous paint layer stratigraphy is crucial in art diagnostics.
• Optically opaque materials impede light penetration, challenging non-invasive analysis.
• Optical coherence tomography (OCT) is a valuable non-invasive tool, but manual analysis is limited.

Purpose of the Study:

• To develop an automated workflow for 3D OCT analysis of paint layers.
• To enable rapid, reproducible, and statistically robust layer thickness quantification.
• To overcome the limitations of manual OCT data analysis in art diagnostics.

Main Methods:

• Utilized supervised semantic segmentation with a Random Forest classifier (Trainable Weka Segmentation).
• Developed a custom MATLAB script for automated interfacial distance calculation.
• Applied the workflow to 3D OCT B-scans for pixel-wise thickness measurements.

Main Results:

• Achieved rapid and fully reproducible pixel-wise thickness measurements from 3D OCT data.
• Demonstrated superior performance compared to manual analysis methods.
• Significantly reduced data processing time for OCT volume analysis.

Conclusions:

• The presented AI-based framework automates OCT segmentation and thickness quantification for cultural heritage research.
• This automated approach offers a significant advancement over traditional manual analysis.
• Enables more efficient and reliable non-invasive characterization of art materials.