Correlations between AI-derived Nanosensor Output and Clinical Diagnosis in Early-Stage Alzheimer’s Disease

Early and accurate diagnosis of Alzheimer’s Disease (AD) is critical for effective management, but current methods are invasive or expensive. Blood-based biosensors offer a promising alternative, yet it is crucial to understand if their objective, quantitative output truly reflects the clinical status of patientsResearch Question: Do changes in objective biomarker concentrations, measured by an AI-powered nanosensor, strongly correlate with the clinical classification (Healthy Control vs. Early AD) of individualsMethods: An electrochemical aptasensor platform was developed to measure Aβ۴۲ and p-tau۱۸۱ in blood plasma. Electrochemical Impedance Spectroscopy (EIS) was used to generate complex impedance data from ۳۰۰ participants (۱۵۰ clinically diagnosed early-AD patients and ۱۵۰ age-matched healthy controls). A Convolutional Neural Network (CNN) was trained to analyze the EIS data and output a continuous “AD Belief Score,” B(AD)B(AD)B(AD), ranging from ۰ (highly probable healthy) to ۱ (highly probable AD). The diagnostic accuracy, sensitivity, and specificity of the classifier were determined. Pearson correlation coefficients were calculated to establish the relationship between the objective B(AD)B(AD)B(AD) score and the discrete clinical diagnosisResults: The AI-powered classifier distinguished between AD patients and healthy controls with ۹۸.۵% accuracy, ۹۹.۱% sensitivity, and ۹۷.۹% specificity. The objective B(AD)B(AD)B(AD) score showed a very strong point-biserial correlation with the clinical diagnosis (rpb=۰.۹۷r_{pb} = ۰.۹۷rpb=۰.۹۷, p<۰.۰۰۱p < ۰.۰۰۱p<۰.۰۰۱). The mean B(AD)B(AD)B(AD) score for the AD cohort was ۰.۹۴±۰.۰۸۰.۹۴ \pm ۰.۰۸۰.۹۴±۰.۰۸, significantly higher than the mean score for the healthy control cohort, which was ۰.۰۵±۰.۰۷۰.۰۵ \pm ۰.۰۷۰.۰۵±۰.۰۷ (p<۰.۰۰۱p < ۰.۰۰۱p<۰.۰۰۱).Significance: In this diagnostic study, the relationship between the objective, sensor-derived score and the established clinical diagnosis is stronger than that reported for many conventional biomarker assays. This suggests that AI-enhanced analysis of complex sensor data captures clinically relevant information with very high fidelity. Such a platform could serve as a reliable, non-invasive tool for large-scale screening and objective tracking of disease status in clinical settings

کلیدواژه ها:

Biosensor ، Electrochemical Impedance Spectroscopy ، Alzheimer’s Disease ، Convolutional Neural Network ، Blood Biomarkers ، Machine Learning ، Aptasensor ، Gold Nanoparticles ، Point-of-Care Testing ، Diagnostics