Application of Machine Learning for the Analysis of Peripheral Blood Biomarkers in Oral Mucosal Diseases: A Cross-Sectional Study

Abstract Background Oral mucosal lesions are widespread globally, yet their pathogenesis remains unclear. Recent evidence suggests that hematological parameters may play a role in their development. This study investigates the differences in humoral immune indexes, micronutrients, and serum vitamin levels between patients with oral mucosal lesions and healthy controls. Additionally, it evaluates the use of a Random Forest machine learning model for classifying various oral mucosal diseases based on peripheral blood biomarkers. Methods We recruited 237 patients with recurrent aphthous ulcers (RAU), 35 with oral lichen planus (OLP), 67 with atrophic glossitis (AG), 35 with burning mouth syndrome (BMS), and 82 healthy controls. Clinical data were recorded using SPSS 24 software. Serum levels of immunoglobulins (IgG, IgA, IgM), complements (C3, C4), vitamins (VB1, VB2, VB3, VB5), serum zinc, serum iron, unsaturated iron-binding capacity (UIBC), total iron-binding capacity (TIBC), and iron saturation were measured and compared across groups. The study also utilized a Random Forest model to analyze a dataset of 319 samples with eight biomarkers. Results Significant differences were found between the patient groups and controls in serum levels of VB2, VB3, VB5, zinc, iron, TIBC, and iron saturation. Levels of VB2 and VB3 were significantly higher in patients compared to controls (p < 0.05), while levels of VB5, serum zinc, serum iron, TIBC, and iron saturation were significantly lower (p < 0.05). No significant differences were observed for C3, C4, IgG, IgM, IgA, VB1, and UIBC. The optimized Random Forest model achieved an accuracy of 94.68% (0.9468) and a Kappa statistic of 0.9306. The model effectively classified certain disease groups, although some overlap was observed. Feature importance analysis identified VB2 (Vitamin B2), VB3 (Vitamin B3), Serum Fe (Serum Iron), TIBC (Total Iron-Binding Capacity), Serum Zn (Serum Zinc)as indicated by Mean Decrease Accuracy and Gini Index. These biomarkers were highlighted as significant based on both the Mean Decrease Accuracy and Mean Decrease Gini Index, indicating their strong contribution to the model’s ability to classify different oral mucosal diseases. Conclusions A strong association was identified between deficiencies in vitamins B2, B3, B5, serum iron, zinc, and other micronutrients and the presence of oral mucosal lesions. Regulation of these vitamin and micronutrient levels may play a crucial role in the prevention and management of such lesions. The model achieved an accuracy of 94.68% (0.9468) and highlighted key biomarkers that significantly contributed to disease classification, demonstrating its potential to enhance our understanding of the pathophysiology and improve diagnostic accuracy for oral mucosal diseases. This shows the utility of machine learning, specifically Random Forest models, for improving the classification and diagnosis of oral mucosal diseases. Future research should aim to validate these findings in larger, independent cohorts and explore other machine learning algorithms to further improve diagnostic accuracy.