CONTRIBUTION OF MAGNETIC RESONANCE IMAGING ASSOCIATED WITH MACHINE LEARNING TO AN EARLY ASSESSMENT OF THE CONVERSION OF MILD COGNITIVE IMPAIRMENT IN OVERT DEMENTIA

Mild Cognitive Impairment (MCI) is a medical condition characterized by noticeable but relatively mild changes in cognitive abilities (memory, thinking, and reasoning) that are greater than expected with normal aging but not severe enough to interfere significantly with daily life or independent function. MCI is broadly classified into amnestic and non-amnestic. Amnestic MCI affects primarily memory and is often considered as a risk factor for Alzheimer’s disease. Non-amnestic MCI affects other thinking skills and may lead to other types of dementia, like frontotemporal dementia or Lewy body dementia. While not all individuals with MCI progress to dementia, identifying those who will convert remains a critical challenge in neurodegenerative research and clinical practice. Magnetic Resonance Imaging (MRI) provides high-resolution structural and functional information about the brain, making it a valuable tool for detecting subtle neuroanatomical changes that precede clinical dementia. However, the complexity and high dimensionality of MRI data limit traditional statistical approaches. Recent advances in machine learning (ML) offer the potential to uncover nonlinear relationships and subtle imaging patterns that can predict disease progression more accurately and earlier than standard diagnostic criteria. We have analyzed MRI findings in 60 subjects aged 76+5 years suffering from amnestic MCI that were examined for 12 months. Subjects were evaluated with a combination of clinical assessment, cognitive testing and imaging tests. Clinical and cognitive testing did not show significant changes along the course of the study. Brain regions analyzed by voxel morphometry included cerebral ventricles, cerebral cortex grey and white matter, temporal cortex, right and left hippocampus and right and left amygdala. Linear mixed models were used to compare within-person rates of change as predictors of MRI biomarkers. Images were pre-processed by skull stripping, normalization, segmentation and structural features were then extracted. This kind of analysis allowed tracking of the rate of brain atrophy over time, earlier than clinical decline and cognitive testing results indicating that MRI combined with ML can enhance early detection of dementia risk among MCI patients.