Integración multimodal de datos cognitivos y morfométricos para el diagnóstico de la etapa de progresión de la enfermedad de Alzheimer

This study addresses the early and accurate diagnosis of Alzheimer's disease, a key challenge in clinical neurology. We investigate an approach that combines morphometric biomarkers from magnetic resonance imaging with cognitive assessments, seeking to overcome the limitations of diagnostic methods that use these information sources in isolation. We analyzed data from 1,382 participants from the ADNI database, classified into cognitive controls, mild cognitive impairment (MCI), and Alzheimer's. We implemented machine learning algorithms, highlighting XGBoost with a balanced accuracy of 83.72%, superior to models based only on cognitive variables. The results indicate that multimodal integration improved the detection of MCI status, with an increase in sensitivity. The biomarkers with the greatest predictive power were hippocampal volumes, entorhinal cortex, and total gray matter measurements. This approach offers a more accurate diagnostic tool and aligns with the NIA-AA research framework, which emphasizes the biological definition of Alzheimer's through biomarkers. An earlier and more precise diagnosis would allow for more timely interventions and better selection of patients for clinical trials. This work demonstrates the potential of artificial intelligence to integrate different sources of information into a predictive model, with possible applications in clinical practice.