Key points are not available for this paper at this time.
Non-invasive characterization of the stratigraphy of heterogeneous paint layers is a key challenge in art diagnostics, due to the presence of optically opaque materials that limit light penetration. Optical coherence tomography (OCT) is widely employed for this purpose thanks to its non-invasiveness, portability, and ease of use. However, conventional layer thickness measurements often rely on manual identification of intensity peaks along individual axial profiles, which is time-consuming, operator-dependent, and impractical for statistically robust analysis of entire OCT volumes. Here, we present an automated workflow for 3D OCT analysis that combines supervised semantic segmentation and post-processing for layer thickness quantification. OCT B-scans are segmented using a Random Forest classifier within the Trainable Weka Segmentation framework to delineate material interfaces, followed by automated calculation of interfacial distances via a custom MATLAB script. This approach enables rapid, fully reproducible, pixel-wise thickness measurements from 3D OCT data, outperforming manual analysis while significantly reducing processing time. To our knowledge, this study is the first application of an AI-based framework for automatic OCT segmentation and thickness quantification in cultural heritage research.
Fovo et al. (Thu,) studied this question.