Innovation
OSCOFAI is critically innovative and highly competitive at the national and international levels, as it combines state-of-the-art AI methods with clinical characteristics and biomarker discovery tools. In detail:
- It includes extensive characterization of biomarkers (MRI, CSF, and blood) that allow us to study various aspects of brain health. Also, this extensive characterization of biomarkers will enable us to study the different aspects of brain health not only in isolation but simultaneously considering potential intercorrelations between them.
- As CSF biomarkers derive from a lumbar puncture, with moderate acceptance from patients, their availability is a great asset of the study, not easily found in population-based studies.
- CSF biomarkers provide a significant window into the brain and have a great diagnostic value for AD.
- Sleep and circadian measurements will not rely on subjective assessment, which is prone to measurement and recall bias, especially in the older age group.
- Sleep and circadian measurements derive from an unobtrusive, objective measure that does not interfere with the everyday life of the participants.
- ALBION study is longitudinal, allowing us to gain better insights on causality. What is more, the study design allows us to focus on the progression of cognitive decline over time, even in the normal spectrum of cognitive function.
- ALBION is a study already up and running with a considerable proportion of the data already available thus, there is high certainty that the project objectives will materialize.
- The use of actigraphy objective data in predicting cognitive decline is currently scarce in the literature. It is limited only to summary measures of whole actigraphy sessions provided by the device vendor’s software. OSCOFAI is the first project proposing to engage in a comprehensive “features engineering” effort aiming to identify new feature sets with high predictive value by analyzing deeper the time series data. We already have encouraging results along this path, to be summarized in the “Preliminary work”.
- It has not been attempted before to combine powerful feature sets with parsimonious ML models for predicting cognitive decline or mediating mechanisms. However, the combination of properly engineered features and simple ML models may give us novel “white box” eXplainable classifiers with the generalization capabilities required for “AI advisor” tools to become helpful in clinical practice in the not so distant future.
- OSCOFAI innovates in that it aims to extract new knowledge using “go-to” proven multivariate linear methods and in addition use AI methods that can exploit it for improved clinical practice.
Impact
Although the population of older people is the most rapidly increasing age group, it remains significantly understudied. This group has to deal with two major problems: age-related cognitive decline and sleep disorders. The increasing prevalence of both problems makes it of great importance to study the potential associations between sleep, measured objectively using actigraphy, with cognitive decline beyond the diagnosis of dementia and mild cognitive impairment, and to reveal the underlying mechanisms and biological connections between sleep and cognition.
- Current research will constitute a cornerstone for investigating various aspects of sleep in relation to cognitive health and will help deliver a prediction tool for cognitive impairment based on objective sleep-wake characteristics.
- The new knowledge and ML/AI predictive tool to be delivered will introduce a relatively unexplored data modality (actigraphy time series) to clinical decision support. Although the actigraphy devices we use in ALBION are medical grade, the new knowledge acquired (in feature engineering, ML pipelines, eXplainable AI, etc.) will be readily transferable to other medical applications using commodity wearable sensors.
- The knowledge obtained will be readily applicable in clinical practice and potentially help older people in their everyday lives. The findings of the current project will be valuable as a basis for clinicians in designing tailored interventions and making specific recommendations for middle-aged and older people under a personalized – precision medicine framework aiming to decelerate cognitive decline.