Multimodal Machine Learning for Early Alzheimer's Disease Prediction: Integrating Biomarkers, Artificial Intelligence, and Precision Neurology

Early and accurate prediction of the onset and progression of Alzheimer’s disease (AD) represents a major medical challenge since years of pathological changes precede the first clinical sign of cognitive decline. So far, clinical prediction of onset and progression of AD has been based on analysis of individual biomarkers or even of data from different modalities. However, the complex and largely unknown biological mechanisms underlying the heterogeneity of AD make it unlikely that a single biomarker or even a single data modality will be sufficient to allow for a robust prediction of disease in individual subjects and of rate of disease progression in already affected subjects. Therefore, there is a strong need for the implementation of machine learning algorithms that integrate in a unifying framework all the different classes of data generated in the context of AD within the framework of precision medicine. Here, we present a framework of multimodal machine learning and discuss its potential for early prediction of AD. We put special emphasis on the potential of the framework for individualized risk assessment, explainable AI, and precision neurology. We also present a series of limitations that may hinder the implementation of the proposed framework in the clinical setting. These limitations are largely due to the heterogeneity of the different classes of data, the lack of large multimodal patient cohorts, the lack of external validation of performance, and the lack of model interpretability. By using models that are at the same time interpretable by humans and validated using clinical endpoints, it may be possible to move towards more individualized methods of early prediction of AD.