Abstract
Artificial intelligence (AI) using fuzzy system offers powerful alternative for modeling complex, nonlinear biomedical relationships, especially under small-sample conditions common in clinical research. This study establishes an fuzzy-driven hybrid framework that combines traditional statistical modeling with fuzzy logic reasoning to predict Alzheimer's disease (AD) risk employing two serum compounds-25-hydroxyvitamin D (25(OH)D) and coenzyme Q10 (CoQ10). Serum samples from 32 AD patients and 32 cognitively healthy controls were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Both levels of the two substances were significantly lower in AD patients compared with controls (p < 0.001). Logistic regression identified 25(OH)D (OR = 0.51) and CoQ10 (OR = 0.45) as independent protective factors, yielding an AUC of 0.86. The fuzzy logic model, based on the Mamdani inference system with nine IF-THEN rules, achieved an accuracy of 81.3% and an AUC of 0.785, indicating stable and explainable performance on a small-scale data. The proposed fuzzy framework effectively bridges quantitative biostatistics and qualitative reasoning, offering a transparent and data-efficient method for biomarker-based AD risk estimation. This explainable AI strategy may provide a good alternative for future applications in accuracy medicine and biomarker-driven disease screening in situations where available are small.