Print Close

Associations between sleep, brain, and cognitive performance using the ENIGMA-Sleep data

Poster No:

2574

Submission Type:

Abstract Submission

Authors:

Hanwen Bi¹^,2, Robin Bülow³, Michele Deantoni⁴, David Elmenhorst⁵^,6^,7, Eva-Maria Elmenhorst⁸^,9, Ralf Ewert¹⁰, Fabio Ferrarelli¹¹, Stefan Frenzel¹², Hans Grabe¹²^,13, Felix Hoffstaedter¹^,2, Neda Jahanshad¹⁴, Ahmadreza Keihani¹¹, Vincent Küppers⁶^,1, Ahmad Mayeli¹¹, Nasrin Mortazavi⁴, Gustav Nilsonne¹⁵^,16, Julia Rupp¹¹, Amin Saberi¹⁷^,1^,2, Christina Schmidt⁴^,18, Kai Spiegelhalder¹⁹, Sandra Tamm¹⁵^,16, Sophia Thomopoulos¹⁴, Paul Thompson¹⁴, Sofie Valk¹⁷^,1^,2, Gilles Vandewalle⁴, Antoine Weihs¹²^,13, Joseph Wexler²⁰^,15, Katharina Wittfeld¹², Simon Eickhoff¹^,2, Kaustubh Patil¹^,2, Federico Raimondo¹^,2, Masoud Tahmasian¹^,2^,6

Institutions:

¹Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich, Jülich, Germany, ²Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf; Heinrich Heine University, Düsseldorf, Germany, ³Institute of Diagnostic Radiology and Neuroradiology, University Medicine Greifswald, Greifswald, Germany, ⁴GIGA-Institute, Cyclotron Research Center/In Vivo Imaging, Sleep and Chronobiology Lab, University of Liège, Liège, Belgium, ⁵Research Center Jülich, Research Center Jülich, Jülich, Germany, ⁶Department of Nuclear Medicine, Faculty of Medicine, University Hospital Cologne, University of Cologne, Cologne, Germany, ⁷Division of Medical Psychology, Rheinische Friedrich-Wilhelms-University Bonn, Bonn, Germany, ⁸Department of Sleep and Human Factors Research, German Aerospace Center, Cologne, Germany, ⁹Institute for Occupational, Social and Environmental Medicine, Medical Faculty, RWTH Aachen University, Aachen, Germany, ¹⁰Internal Medicine B, Pneumology, University Medicine Greifswald, Greifswald, Germany, ¹¹Department of Psychiatry, University of Pittsburgh, Pittsburgh, PA, USA, ¹²Department of Psychiatry and Psychotherapy, University Medicine Greifswald, Greifswald, Germany, ¹³German Center for Neurodegenerative Diseases (DZNE), Greifswald, Germany, ¹⁴Imaging Genetics Center, Keck School of Medicine of University of Southern California, Los Angeles, CA, USA, ¹⁵Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden, ¹⁶Department of Psychology, Stress Research Institute, Stockholm University, Stockholm, Sweden, ¹⁷Otto Hahn Research Group for Cognitive Neurogenetics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, ¹⁸Psychology and Neuroscience of Cognition Research Unit, Faculty of Psychology and Educational Sciences, University of Liège, Liège, Belgium, ¹⁹Department of Psychiatry and Psychotherapy, Medical Center - University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg im Breisgau, Germany, ²⁰Department of Psychology, Stanford University, Stanford, CA, USA

First Author:

Hanwen Bi
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf; Heinrich Heine University
Jülich, Germany|Düsseldorf, Germany

Co-Author(s):

Robin Bülow
Institute of Diagnostic Radiology and Neuroradiology, University Medicine Greifswald
Greifswald, Germany

Michele Deantoni
GIGA-Institute, Cyclotron Research Center/In Vivo Imaging, Sleep and Chronobiology Lab, University of Liège
Liège, Belgium

David Elmenhorst
Research Center Jülich, Research Center Jülich|Department of Nuclear Medicine, Faculty of Medicine, University Hospital Cologne, University of Cologne|Division of Medical Psychology, Rheinische Friedrich-Wilhelms-University Bonn
Jülich, Germany|Cologne, Germany|Bonn, Germany

Eva-Maria Elmenhorst
Department of Sleep and Human Factors Research, German Aerospace Center|Institute for Occupational, Social and Environmental Medicine, Medical Faculty, RWTH Aachen University
Cologne, Germany|Aachen, Germany

Ralf Ewert
Internal Medicine B, Pneumology, University Medicine Greifswald
Greifswald, Germany

Fabio Ferrarelli
Department of Psychiatry, University of Pittsburgh
Pittsburgh, PA, USA

Stefan Frenzel
Department of Psychiatry and Psychotherapy, University Medicine Greifswald
Greifswald, Germany

Hans Grabe
Department of Psychiatry and Psychotherapy, University Medicine Greifswald|German Center for Neurodegenerative Diseases (DZNE)
Greifswald, Germany|Greifswald, Germany

Felix Hoffstaedter
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf; Heinrich Heine University
Jülich, Germany|Düsseldorf, Germany

Neda Jahanshad, PhD
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, CA, USA

Ahmadreza Keihani
Department of Psychiatry, University of Pittsburgh
Pittsburgh, PA, USA

Vincent Küppers
Department of Nuclear Medicine, Faculty of Medicine, University Hospital Cologne, University of Cologne|Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich
Cologne, Germany|Jülich, Germany

Ahmad Mayeli
Department of Psychiatry, University of Pittsburgh
Pittsburgh, PA, USA

Nasrin Mortazavi
GIGA-Institute, Cyclotron Research Center/In Vivo Imaging, Sleep and Chronobiology Lab, University of Liège
Liège, Belgium

Gustav Nilsonne
Department of Clinical Neuroscience, Karolinska Institutet|Department of Psychology, Stress Research Institute, Stockholm University
Stockholm, Sweden|Stockholm, Sweden

Julia Rupp
Department of Psychiatry, University of Pittsburgh
Pittsburgh, PA, USA

Amin Saberi
Otto Hahn Research Group for Cognitive Neurogenetics, Max Planck Institute for Human Cognitive and Brain Sciences|Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf; Heinrich Heine University
Leipzig, Germany|Jülich, Germany|Düsseldorf, Germany

Christina Schmidt
GIGA-Institute, Cyclotron Research Center/In Vivo Imaging, Sleep and Chronobiology Lab, University of Liège|Psychology and Neuroscience of Cognition Research Unit, Faculty of Psychology and Educational Sciences, University of Liège
Liège, Belgium|Liège, Belgium

Kai Spiegelhalder
Department of Psychiatry and Psychotherapy, Medical Center - University of Freiburg, Faculty of Medicine, University of Freiburg
Freiburg im Breisgau, Germany

Sandra Tamm
Department of Clinical Neuroscience, Karolinska Institutet|Department of Psychology, Stress Research Institute, Stockholm University
Stockholm, Sweden|Stockholm, Sweden

Sophia Thomopoulos
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, CA, USA

Paul Thompson, PhD
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, CA, USA

Sofie Valk
Otto Hahn Research Group for Cognitive Neurogenetics, Max Planck Institute for Human Cognitive and Brain Sciences|Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf; Heinrich Heine University
Leipzig, Germany|Jülich, Germany|Düsseldorf, Germany

Gilles Vandewalle
GIGA-Institute, Cyclotron Research Center/In Vivo Imaging, Sleep and Chronobiology Lab, University of Liège
Liège, Belgium

Antoine Weihs
Department of Psychiatry and Psychotherapy, University Medicine Greifswald|German Center for Neurodegenerative Diseases (DZNE)
Greifswald, Germany|Greifswald, Germany

Joseph Wexler
Department of Psychology, Stanford University|Department of Clinical Neuroscience, Karolinska Institutet
Stanford, CA, USA|Stockholm, Sweden

Katharina Wittfeld
Department of Psychiatry and Psychotherapy, University Medicine Greifswald
Greifswald, Germany

Simon Eickhoff
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf; Heinrich Heine University
Jülich, Germany|Düsseldorf, Germany

Kaustubh Patil
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf; Heinrich Heine University
Jülich, Germany|Düsseldorf, Germany

Federico Raimondo
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf; Heinrich Heine University
Jülich, Germany|Düsseldorf, Germany

Masoud Tahmasian
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf; Heinrich Heine University|Department of Nuclear Medicine, Faculty of Medicine, University Hospital Cologne, University of Cologne
Jülich, Germany|Düsseldorf, Germany|Cologne, Germany

Introduction:

Sleep disturbances potentially exacerbate cognitive decline and dementia[1],[2], necessitating large-scale studies to explore the neurobiological interplay between sleep and cognitive performance. While previous studies highlighted a non-linear relationship between sleep duration and cognitive performance, their effect size was small[3],[4], they used group comparison statistical approaches, focused on subjective sleep assessment, and were limited to a particular sample, limiting the generalizability of their findings in other samples. Machine learning (ML) methods enable individual-level predictions and can validate models on unseen data, providing a more robust analytical framework. Here, we performed ML analysis based on objective and subjective sleep measurements and brain structure data from the ENIGMA-Sleep working group[5] to predict cognitive performance.

Methods:

A total of 1,023 participants from SHIP-Trend[6] Greifswald (N=831) and Liege[7] (N=192) have been included. Sleep measurements were sleep duration and sleep efficiency extracted from Polysomnography data and Pittsburgh Sleep Quality Index. Cognitive performance was measured using Stroop scores (SHIP-Trend and Liege) and N-back working memory (Liege). Subcortical volumes derived from the Aseg atlas via FreeSurfer for brain structure data[8]. In addition, age, sex, BMI, and depressive symptoms (Beck Depression Inventory-II) scores were added as model input as demographic/behavioral factors. Three machine learning models were utilized in this study: the support vector machine (SVM) with RBF kernel, random forest (RF), and AutoGluon[9]. These models were trained on the SHIP-Trend and Liege datasets separately using nested cross-validation, with the performance metrics being derived from the average results across this process (Figure 1A). Out-of-sample predictions were conducted by applying the models, which were trained on the SHIP-Trend dataset, to the Liege dataset (Figure 1B). The performance of ML models was evaluated by mean absolute error, mean squared error, R2, and Pearson correlation. Different feature combinations of sleep measurement, subcortical volume, and demographic/behavioral factors were used as model inputs. SHAP toolbox[10] was utilized for model explanation, elucidating the importance and influence of various features on the predictions.

·Machine learning pipeline for cognitive performance prediction in SHIP-Trend and Liege datasets

Results:

Sleep measurements and covariates predicted Stroop reaction times in SHIP-Trend (AutoGluon: R2=0.1308, Pearson correlation=0.3812) and interference scores in Liege (AutoGluon: R2=0.2716, Pearson correlation=0.5593) using SVM, RF, and AutoGluon (Figure 2A). The best performance was achieved when using sleep measurement, subcortical volumes, and covariates together in the SHIP-Trend dataset (AutoGluon: R2=0.1409, Pearson correlation=0.3959) (Figure 2B). In the out-of-sample validation using the AutoGluon model, although there was a strong correlation between the predicted values and the actual Stroop interference scores in the Liege dataset (Pearson correlation=0.4186) (Figure 2C), the R2 was adversely affected due to the differing scales of Stroop interference scores between the SHIP-Trend and Liege datasets. Across all models and both datasets, age emerged as a key predictive feature (Figure 2D). However, in the Liege dataset, the models were unable to predict working memory accuracy.

·Machine learning results

Conclusions:

Although the predictive strength is modest, non-linear ML models have a consistent and stable capacity to predict Stroop interference and reaction time scores using our input features across both samples. The SHAP findings showed that age was the most important contributing factor to our predictive model. Particularly, older individuals exhibiting shorter sleep durations and younger individuals with longer sleep durations contribute to prediction, indicating the complex relationships between sleep, brain, and aging to predict cognitive function at the individual subject level.

Lifespan Development:

Aging

Modeling and Analysis Methods:

Classification and Predictive Modeling ²

Neuroanatomy, Physiology, Metabolism and Neurotransmission:

Subcortical Structures

Perception, Attention and Motor Behavior:

Sleep and Wakefulness ¹

Keywords:

Aging

Cognition

Machine Learning

Sleep

STRUCTURAL MRI

^1|2Indicates the priority used for review

Provide references using author date format

1. Emamian, F. et al. The Association Between Obstructive Sleep Apnea and Alzheimer’s Disease: A Meta-Analysis Perspective. Frontiers in Aging Neuroscience 8, (2016).
2. Koren, T., Fisher, E., Webster, L., Livingston, G. & Rapaport, P. Prevalence of sleep disturbances in people with dementia living in the community: A systematic review and meta-analysis. Ageing Research Reviews 83, 101782 (2023).
3. Kyle, S. D. et al. Sleep and cognitive performance: cross-sectional associations in the UK Biobank. Sleep Medicine 38, 85–91 (2017).
4. Li, Y. et al. The brain structure and genetic mechanisms underlying the nonlinear association between sleep duration, cognition and mental health. Nat Aging 2, 425–437 (2022).
5. Tahmasian, M. et al. ENIGMA‐Sleep: Challenges, opportunities, and the road map. J. Sleep Res. 30, (2021).
6. Völzke, H. et al. Cohort Profile Update: The Study of Health in Pomerania (SHIP). International Journal of Epidemiology 51, e372–e383 (2022).
7. Deantoni, M. et al. Association between sleep slow-wave activity and in-vivo estimates of myelin in healthy young men. NeuroImage 272, 120045 (2023).
8. Fischl, B. et al. Whole Brain Segmentation: Automated Labeling of Neuroanatomical Structures in the Human Brain. Neuron 33, 341–355 (2002).
9. Erickson, N. et al. AutoGluon-Tabular: Robust and Accurate AutoML for Structured Data. Preprint at http://arxiv.org/abs/2003.06505 (2020).
10. Lundberg, S. M. & Lee, S.-I. A unified approach to interpreting model predictions. in Proceedings of the 31st International Conference on Neural Information Processing Systems 4768–4777 (Curran Associates Inc., 2017).