Kuo et al.), however using more layers in these CNNs reaching an MAE of 3.33 years.ĭeep learning models performed better than classic machine learning algorithms such as Support Vector Machines, Relevance Vector Machines or Gaussian Process Regression only one of which reached an MAE of 3.09 years being submitted but not published. The third and fourth place finishing teams also used an ensemble of differently preprocessed data and 3D CNN architectures (see Couvy-Duchesne et al. This approach reached the lowest mean absolute errors (MAE) of 2.90 years and 2.95 years for the first and second objective of the PAC, respectively. used lightweight 3D Convolutional Neural Networks (CNNs) based on the Fully Convolutional Network ( 4) and the VGG net ( 5) combined with different preprocessing steps to form a multimodal ensemble which was pre-trained on N = 14,503 samples from UK Biobank. In this special issue, 8 publications including the top performing models of the PAC 2019, are described in detail to provide insights into cutting-edge brain-age modeling and an outlook on future developments. The PAC 2019 featured 274 participants in 79 teams from across the globe and resulted in a great variety of submitted machine learning models. The second objective was to minimize the brain-age gap while keeping the Spearman correlation between the brain-age gap and chronological age below r = 0.10 to avoid the commonly observed bias of age estimations toward the mean age of the training dataset, as discussed by Treder et al. Based on a large structural Magnetic Resonance Imaging (sMRI) dataset of N = 3,307 healthy individuals (see Table 1) provided by the two organizing sites (University of Münster, King's College London), the PAC 2019 pursued two goals: first, participants aimed to minimize the mean difference between chronological and predicted age (i.e., the brain-age gap). The Predictive Analytics Competition (PAC) 2019 aimed to bring together machine learning and neuroimaging experts to improve existing brain-age models. This trained model is then used to evaluate neuroimaging data from previously unseen individuals and evaluated based on the brain-age gap as defined by the difference between predicted and chronological age. In a typical brain-age study, a machine learning model is trained on neuroimaging data-usually whole-brain structural T 1-weighted Magnetic Resonance Imaging (MRI) data-to predict chronological age. Building on this, the so-called brain-age paradigm ( 1) aims to estimate a brain's biological age ( 2) and may serve as a cumulative marker of disease-risk, functional capacity and residual lifespan ( 3). Though aging is ubiquitous, the rate at which age-associated biological changes in the brain occur differs substantially between individuals.
0 kommentar(er)
