Automated and highly precise surface wetting contact angle measurement with optical coherence tomography based on deep learning model

Accurately determining the contact angle (CA) is critical for analyzing the wetting properties of materials and investigating solid–liquid interactions. This study presents a novel approach for predicting the CA of liquid droplets on three distinct material surfaces, High-Density Polyethylene (HDPE), Polystyrene (PS), and Polytetrafluoroethylene (PTFE), using Optical Coherence Tomography (OCT) due to providing high-resolution, non-contact, and three-dimensional structural imaging. We created a dataset from volumetric OCT images and then, developed and comprehensively evaluated machine learning and deep learning models, leveraging deep features extracted from five variations of the Next Generation of Convolutional Networks (ConvNeXt) architecture to enhance CA prediction accuracy. The extracted deep features were applied to both traditional machine learning (ML) models, such as Random Forest and Support Vector Regression, and advanced deep learning (DL) models, including Long Short-Term Memory (LSTM) and Bi-directional LSTM (Bi-LSTM). Results reveal that DL models, particularly the Bi-LSTM with ConvNeXt-Tiny features, consistently outperformed classical ML models across all material types. This model achieved the highest predictive accuracy, with superior R² values, reduced error rates, and strong consistency, as validated by regression fitting and Bland-Altman analyses. These findings highlight the robustness and versatility of the proposed study capturing volumetric OCT images and the DL framework for material-independent CA prediction, with potential implications for advancing surface wettability research and applications in a wide range such as coating technologies, material design, or biomedical surface analysis.