Longitudinal Multi‐Domain Biomarkers and Machine Learning for Predicting Alzheimer's Disease Progression

Figure 1 reveals significant trends across cognitive, memory, and imaging biomarkers. Cognitive measures (e.g., CDRSB_change, ADAS11_change) increased with severity, while memory measures (e.g., RAVLT_immediate_change) declined, with MCI-to-AD transitions showing the steepest changes. Imaging biomarkers, such as Ventricles_change and Hippocampus_change, indicated marked structural atrophy in transitions to AD. The Random Forest Classifier achieved strong performance, with AUC values of 0.76-0.98 (Figure 2), excelling in MCI-to-AD transitions. Sensitivity-specificity analyses (Figure 3) highlighted a trade-off, showcasing the model's ability to distinguish stable stages from transitions effectively CONCLUSION: This study demonstrates the effectiveness of machine learning with multi-domain biomarkers in predicting Alzheimer's disease progression. The Random Forest Classifier achieved high AUC values, with key biomarkers like CDRSB_change and Hippocampus_change identifying transitions. This approach supports early interventions and personalized care, with future work focusing on longer follow-ups and additional biomarkers.