Gast ::
| Hauptseite > Publikationsdatenbank > Prediction of individual cognitive test performance based on imaging and non-imaging data in older adults |
| Hauptseite > Publikationsdatenbank > Prediction of individual cognitive test performance based on imaging and non-imaging data in older adults |
| Conference Presentation (After Call) | FZJ-2025-05034 |
; ; ; ; ;
2025
Abstract: Early detection of cognitive decline gains relevance in normal aging given its impact on the quality of life of older adults. While using brain imaging data alone can be challenging, there is an opportunity to use health-related and demographic data as biomarker as these are easily accessible and have already been shown to be associated with cognitive dysfunction.Thus, using machine learning (ML) we examined the practicality of 1) imaging, 2) health related and 3) demographic data, in the prediction of cognitive functioning (16 cognitive test scores) in 494 older adults (67 +/- 7 years) from 1000BRAINS. Prediction performance was obtained for each modality and its combinations using cross-validation and four algorithms.Predictability differences emerged across modalities and cognitive functions. In terms of individual tests, vocabulary, executive and episodic memory functions were moderately predicted from demographic and partially from brain data; working memory showed low predictability across modalities.Overall, health-related data showed limited predictability across cognitive functions despite known associations between cardiovascular health and cognitive decline. Strikingly, demographic variables outperformed health and imaging data highlighting their impact on predictions of cognition. Finally, we observed higher predictability of executive and episodic memory functions, which are important for the prognosis of neurodegenerative diseases.
|
The record appears in these collections: |