Dementia and cognitive impairment due to Alzheimer's disease (AD), are common in the elderly population worldwide. Diagnosis of AD is based on clinical evaluation, neuropsychological assessment tools, biomarkers and imaging modalities such as brain Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) scan. Since anti-amyloid monoclonal antibodies for AD have been recently introduced, early diagnosis is essential for timely treatment initiation. To address diagnostic needs and monitor treatment related side effects, increasingly sophisticated imaging techniques have been developed and integrated into clinical practice. Despite these advancements, effectively integrating multimodal data across several timepoints continues to be a major obstacle in clinical practice. In recent years, Artificial intelligence (AI) has benefited healthcare in many ways, through automated decision-making, diagnosis, patient monitoring and more. In the field of dementia and AD specifically, the application of AI tools to data collected through neuroimaging leads to significant advances in diagnostic process and patients' surveillance. The current narrative review examines recent studies that apply deep learning methods to commonly used diagnostic neuroimaging techniques, focusing on the early stages of AD and its diagnosis. We highlight both the advantages and limitations of these approaches, thereby providing an overview of current AI implementation in diagnostic neuroimaging. We also suggest a direction for future research to overcome the limitations of currently employed tools and further advance early diagnosis and treatment of AD.
Lee et al. (Wed,) studied this question.