Machine learning-based electrical impedance spectroscopy classification for oral cancer

Head and neck cancers, particularly oral squamous cell carcinoma (OSCC), are major global health concern. An early detection can significantly improve the survival rate. Traditional diagnostic methods, such as biopsy, are invasive and analysis subjective, posing challenges for routine screening. This study explores the application of electrical impedance spectroscopy (EIS) combined with machine learning (ML) to classify oral cancer and healthy tissues non-invasively. By utilizing tissue-engineered oral mucosal model-based finite element modeling (FE) and simulations, we generate synthetic EIS data to train ML classifiers and apply the classifiers to real tissue data to classify the tissue status, Among the ML models tested, Random Forest (RF) showed the best performance, achieving an AVC of 0.95 and good sensitivity and specificity. This approach overcomes data scarcity issues by leveraging synthetic data and provides a promising pathway toward reliable, non-invasive diagnostic tools for oral cancer. Further research will focus on enhancing model accuracy by integrating clinical prior knowledge and refining tissue engineering techniques.