Assimetrias inter-hemisféricas no hipocampo e sua associação com a progressão da doença de Alzheimer

Abstract

Brain atrophy is a progressive process typically associated with natural aging. In neurodegenerative diseases such as Alzheimer’s disease (AD), this process is accelerated, manifesting in symptoms related to memory loss, reasoning difficulties, and behavioral changes. Currently, there is no cure for AD, and early treatment remains the most effective way to mitigate or delay symptoms. In addition to clinical tests and family history, structural Magnetic Resonance Imaging (MRI) of the brain plays a fundamental role in diagnosis, as it enables visual assessment of brain atrophy—allowing comparison with age-related atrophy—and assists in ruling out other possible comorbidities from treatable dementias. Moreover, inter-hemispheric alterations are associated with AD and can be observed in MRI scans. Recent studies have investigated new methods for quantifying such asymmetries in order to develop biomarkers for supporting AD diagnosis and to train classifiers for automated diagnosis. However, despite promising results, visual inspection remains the most common approach for evaluating AD presence in MRI images. In this context, this work proposes a new methodology to quantify inter-hemispheric hippocampal asymmetries based on geometric and textural descriptors, focusing on predicting the progression from stable to progressive Mild Cognitive Impairment (MCI). The approach employs the shape context technique to extract log-spherical histograms from the accumulated energy map of the 3D Dual-Tree Complex Wavelet Transform (DT-CWT) in the hippocampal region. These descriptors, combined with volumetric features, are fed into a Convolutional Siamese neural network specifically designed to model inter-hemispheric differences. The proposed methodology achieved an F1-score of 0.845 ± 0.01 in distinguishing stable from progressive MCI cases. Additional binary classification experiments involving cognitively normal (CN), MCI, and AD subjects were also conducted, with consistent results: CN vs. AD (F1-score = 0.872 ± 0.05), CN vs. MCI (0.757 ± 0.03), and MCI vs. AD (0.731 ± 0.05). These findings demonstrate the influence of inter-hemispheric hippocampal alterations during AD progression and the effectiveness of these features for disease prediction and prognosis.