This study evaluates the effectiveness of support vector machines (SVMs) in classifying structural MRI scans for Alzheimer's disease (AD) and distinguishing it from normal aging. The research aims to assess the accuracy of SVMs in assigning individual diagnoses and to determine if data from multiple scanners and centers can be combined to achieve effective classification. The study used linear SVMs to classify T1-weighted MRI scans of AD patients and cognitively normal elderly individuals from two centers with different scanning equipment. The results show that up to 96% of pathologically verified AD patients were correctly classified using whole-brain images. Data from different centers were successfully combined, achieving comparable results. Importantly, data from one center could be used to train a SVM to accurately differentiate AD and normal aging scans obtained from another center. The method also performed well in differentiating mild AD from controls and between AD and frontotemporal lobar degeneration (FTLD). The study concludes that SVMs successfully separate AD patients from healthy aging subjects, perform well in differential diagnosis of different forms of dementia, and are robust and generalizable across different centers, suggesting a potential role for computer-based diagnostic image analysis in clinical practice.This study evaluates the effectiveness of support vector machines (SVMs) in classifying structural MRI scans for Alzheimer's disease (AD) and distinguishing it from normal aging. The research aims to assess the accuracy of SVMs in assigning individual diagnoses and to determine if data from multiple scanners and centers can be combined to achieve effective classification. The study used linear SVMs to classify T1-weighted MRI scans of AD patients and cognitively normal elderly individuals from two centers with different scanning equipment. The results show that up to 96% of pathologically verified AD patients were correctly classified using whole-brain images. Data from different centers were successfully combined, achieving comparable results. Importantly, data from one center could be used to train a SVM to accurately differentiate AD and normal aging scans obtained from another center. The method also performed well in differentiating mild AD from controls and between AD and frontotemporal lobar degeneration (FTLD). The study concludes that SVMs successfully separate AD patients from healthy aging subjects, perform well in differential diagnosis of different forms of dementia, and are robust and generalizable across different centers, suggesting a potential role for computer-based diagnostic image analysis in clinical practice.