Development of Neuroimaging-based Biomarkers for Early Detection of Alzheimer's Disease

Alzheimer's Disease (AD), the leading cause of dementia, is characterized by a progressive cognitive decline and structural brain degeneration. This study is aimed to identify MRI-based biomarkers for early diagnosis of AD, by analyzing data from three clinical groups: controls, mild cognitive impairment (MCI) and AD patients. A total of 589 parameters were extracted: 536 from volumetry, 35 from diffusion, and 12 from perfusion measurements. Statistical analysis focused on two classification tasks: a binary classification to differentiate between controls and unhealthy patients, and a multiclass classification to distinguish between controls, MCI and AD patients. Key volumetric parameters included gray and white matter, hippocampus, limbic system, ventricles and the subventricular zone (SVZ). All the regions analyzed for diffusion and perfusion provided at least one significant parameter. Thirty key biomarkers were selected for each classification task and used to train machine learning models. These were KNN, SVM and Ensemble methods, trained with 5, 10, 15, 20, 25 and 30 parameters as well as a PCA 95%. Ensemble models achieved the highest accuracy, with 85% for binary classification and 75% for multiclass classification. These findings demonstrate the promise of MRI biomarkers combined with machine learning techniques for enhancing early diagnosis of AD.Clinical Relevance- This project provides innovative MRI-based biomarkers for Alzheimer's diagnosis. Early identification of neurodegeneration is crucial for improving patient outcomes. These findings could also influence on diagnostic protocols to enable a more accurate differentiation between healthy individuals, those with mild cognitive impairment and Alzheimer's Disease patients.