Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that affects millions of people worldwide. Detecting and predicting the progression of AD from mild cognitive impairment (MCI) to AD dementia is crucial for early intervention and treatment. In a recent systematic review published in BMC Neurology, researchers examined the use of neuroimaging techniques combined with advanced machine learning algorithms for predicting the progression of AD.
Neuroimaging modalities such as positron emission tomography (PET), single-photon emission computed tomography (SPECT), MRI, functional MRI (fMRI), and electroencephalogram (EEG) can provide valuable information about the structural and functional changes in the brain associated with AD. However, integrating and analyzing data from different modalities pose challenges.
Machine learning (ML) methods and deep learning algorithms offer a solution for analyzing high-dimensional neuroimaging data to identify individuals at risk of developing AD dementia. These approaches can help create predictive models that aid in the early detection of AD and facilitate targeted interventions for individuals at high risk of progression.
One challenge in developing ML-based models for AD prediction is the requirement for large training samples, while available AD datasets often have limited sample sizes. Researchers followed the PRISMA guidelines for their systematic review, searching three electronic databases for recent scientific work on AD prediction. They included studies that used neuroimaging modalities, described the methodology in detail, and reported accuracy measures for comparison.
The majority of studies screened MCI patients using structural MRI and PET scans, but these modalities have limitations in terms of cost, maintenance, and mobility. EEG, on the other hand, is affordable and easy to implement, making it a promising modality for MCI screening. However, the lack of longitudinal EEG datasets hinders its use in AD progression studies.
The researchers also found that most studies used simple linear methods for classification and relied on relatively short follow-up periods. Collaborative efforts between computer science, neuroscience, and cognitive science researchers are needed to overcome these challenges and develop more accurate and practical predictive models for AD progression.
In conclusion, the field of neuroimaging and machine learning for AD prediction is advancing rapidly. Future research should focus on developing deep learning approaches that can analyze brain scans with high precision and address the limitations of current neuroimaging modalities. These advancements could lead to the early detection and targeted treatment of AD, offering hope for those at risk of developing AD dementia.