Automated brain-age prediction and its interpretation

• Applicants: Professor Dr. Simon B. Eickhoff; Kaustubh R. Patil, Ph.D.
• Subject Area: Cognitive, Systems and Behavioural Neurobiology
• Term: from 2019 to 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 432015680

• Final Report Year: 2024

Final Report Abstract

The biological age of a brain, or brain-age, is a key indicator of its structural or functional state. Notably, an elevated brain-age, reflected in the BrainAGE score, is a biomarker for identifying individuals at heightened risk of age-related diseases. However, brain-age is not directly measurable and must be estimated, necessitating precise methods using non-invasive brain imaging data. While various machine learning (ML) techniques have been proposed for predicting brain-age, two areas remain significantly underexplored. First, the influence of manifold choices in data selection, representation, ML algorithms, and their interaction with specific contexts (like limited age ranges or particular sites) on prediction accuracy has not been thoroughly assessed. As no single method excels in every scenario, identifying effective workflows—that combine data representation, ML algorithms, and context—is essential to enhance prediction accuracy. Second, identifying age-sensitive brain regions that contribute to deviations in brain-age from chronological age has been only minimally examined. Additionally, the role of individual-specific factors, such as atypical aging in neurodegenerative diseases, is not well understood. This study aims to tackle these challenges by (1) offering workflow design guidelines through systematic evaluation of numerous workflows and contexts within a big-data framework, and (2) enhancing neurobiological understanding by developing and applying interpretation methods to pinpoint age-sensitive brain areas and clarify individual predictions.

Publications

• ECML-PCKDD 2020: Confound Removal and Normalization in Practice: A Neuroimaging Based Sex Prediction Case Study
  Shammi More
  
• Advanced brain ageing in Parkinson’s disease is related to disease duration and individual impairment. Brain Communications, 3(3).
  Eickhoff, Claudia R.; Hoffstaedter, Felix; Caspers, Julian; Reetz, Kathrin; Mathys, Christian; Dogan, Imis; Amunts, Katrin; Schnitzler, Alfons & Eickhoff, Simon B.
  
• Confound Removal and Normalization in Practice: A Neuroimaging Based Sex Prediction Case Study. Lecture Notes in Computer Science, 3-18. Springer International Publishing.
  More, Shammi; Eickhoff, Simon B.; Caspers, Julian & Patil, Kaustubh R.
  
• FZJ end of year lecture 2021: The Organization of Our Brain: Tracking Aging and Mental Illness with Big Data and AI
  Kaustubh Patil
  
• HIDA Annual Conference 2021 (Scientific Telegram Session): Brainage prediction: a systematic comparison of machine learning workflows
  Shammi More
  
• INM&IBI Retreat 2021, A comparison of VBM pipelines using large structural MRI datasets
  G. Antonopoulos, F. Hoffstaedter, F. Raimondo, S. B. Eickhoff & K. R. Patil
  
• INM-IBI Retreat 2021: Brain-age prediction: a systematic comparison of machine learning workflows
  Shammi More
  
• OHBM Annual meeting 2021: A comparison of VBM pipelines
  G. Antonopoulos, F. Hoffstaedter, F. Raimondo, S. B. Eickhoff & K. R. Patil
  
• INM-IBI Retreat 2022: Brain-age prediction: a systematic comparison of machine learning workflows
  Georgios Antonopoulos
  
• OHBM Annual conference 2022: Brain-age prediction: a systematic comparison of machine learning workflows
  Shammi More
  
• Reporting details of neuroimaging studies on individual traits prediction: A literature survey. NeuroImage, 256, 119275.
  Yeung, Andy Wai Kan; More, Shammi; Wu, Jianxiao & Eickhoff, Simon B.
  
• A systematic comparison of VBM pipelines and their application to age prediction. NeuroImage, 279, 120292.
  Antonopoulos, Georgios; More, Shammi; Raimondo, Federico; Eickhoff, Simon B.; Hoffstaedter, Felix & Patil, Kaustubh R.
  
• Brain-age prediction: A systematic comparison of machine learning workflows. NeuroImage, 270, 119947.
  More, Shammi; Antonopoulos, Georgios; Hoffstaedter, Felix; Caspers, Julian; Eickhoff, Simon B. & Patil, Kaustubh R.
  
• Helmholtz AI 2023: Stacking ensemble for age-prediction improves performance and privacy
  G. Antonopoulos, S. More, F. Raimondo & K. R. Patil
  
• MRI or18F-FDG PET for Brain Age Gap Estimation: Links to Cognition, Pathology, and Alzheimer Disease Progression. Journal of Nuclear Medicine, 65(1), 147-155.
  Doering, Elena; Antonopoulos, Georgios; Hoenig, Merle; van Eimeren, Thilo; Daamen, Marcel; Boecker, Henning; Jessen, Frank; Düzel, Emrah; Eickhoff, Simon; Patil, Kaustubh & Drzezga, Alexander
  
• Julearn: an easy-to-use library for leakage-free evaluation and inspection of ML models. Gigabyte, 2024, 1-16.
  Hamdan, Sami; More, Shammi; Sasse, Leonard; Komeyer, Vera; Patil, Kaustubh R. & Raimondo, Federico