Deep Neural Networks and Kernel Regression Achieve Comparable Accuracies for Functional Connectivity Prediction of Behavior and Demographics

He, Tong; Kong, Ru; Holmes, Avram J.; Nguyen, Minh; Sabuncu, Mert R.; Eickhoff, Simon B.; Bzdok, Danilo; Feng, Jiashi; Yeo, B.T. Thomas

doi:10.1101/473603

Cited by 68 publications

(127 citation statements)

References 112 publications

Supporting

Mentioning

116

Contrasting

Order By: Relevance

“…Interestingly, the top team considered additional handcrafted features, which might have contributed to its success. Furthermore, the top team utilized a non-deep-learning algorithm XGboost (Chen and Guestrin, 2016), which might be consistent with recent work suggesting that for certain neuroimaging applications, non-deep-learning approaches might be highly competitive (He et al, 2019)…”

Section: Discussionsupporting

confidence: 78%

Predicting Alzheimer’s disease progression using deep recurrent neural networks

Nguyen

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

Early identification of individuals at risk of developing Alzheimer's disease (AD) dementia is important for developing disease-modifying therapies. In this study, given multimodal AD markers and clinical diagnosis of an individual from one or more timepoints, we seek to predict the clinical diagnosis, cognition and ventricular volume of the individual for every month (indefinitely) into the future. We proposed a recurrent neural network (RNN) model and applied it to data from The Alzheimer's Disease Prediction Of Longitudinal Evolution (TADPOLE) challenge, comprising longitudinal data of 1677 participants (Marinescu et al. 2018) from the Alzheimer's Disease Neuroimaging Initiative (ADNI). We compared the performance of the RNN model and three baseline algorithms up to 6 years into the future.Most previous work on predicting AD progression ignore the issue of missing data, which is a prevalent issue in longitudinal data. Here, we explored three different strategies to handle missing data. Two of the strategies treated the missing data as a "preprocessing" issue, by imputing the missing data using the previous timepoint ("forward filling") or linear interpolation ("linear filling). The third strategy utilized the RNN model itself to fill in the missing data both during training and testing ("model filling"). Our analyses suggest that the RNN with "model filling" was better than baseline algorithms, including support vector machine/regression and linear state space (LSS) models. However, there was no statistical difference between the RNN and LSS for predicting cognition and ventricular volume.Importantly, although the training procedure utilized longitudinal data, we found that the trained RNN model exhibited similar performance, when using only 1 input timepoint or 4 input timepoints, suggesting that our approach might work well with just cross-sectional data.An earlier version of our approach was ranked 5th (out of 53 entries) in the TADPOLE challenge in 2019. The current approach is ranked 2nd out of 56 entries as of August 12th, 2019.

show abstract

Section: Discussionsupporting

confidence: 78%

Predicting Alzheimer’s disease progression using deep recurrent neural networks

Nguyen

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The topography of these population-based network solutions are closely coupled to cognitive function, and a strong correspondence has been observed linking the spatial structure of intrinsic (fcMRI) and extrinsic (task-evoked) networks of the human brain [21][22][23] . Consistent with these observations, various connectivity patterns track behavioral variability in the general population [24][25][26] and symptom expression in patients with psychiatric illness 27 . Suggesting genetic factors may influence the functioning of large-scale brain networks, patterns of intrinsic connectivity within population-average defined network templates are heritable [28][29][30] and act as a trait-like fingerprint that can accurately identify specific people from a larger group 31,32 .…”

Section: Introductionsupporting

confidence: 66%

Heritability of individualized cortical network topography

Anderson

Kong

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Human cortex is patterned by a complex and interdigitated web of large-scale functional networks. Recent methodological breakthroughs reveal variation in the size, shape, and spatial topography of cortical networks across individuals. While spatial network organization emerges across development, is stable over time, and predictive of behavior, it is not yet clear to what extent genetic factors underlie inter-individual differences in network topography. Here, leveraging a novel non-linear multi-dimensional estimation of heritability, we provide evidence that individual variability in the size and topographic organization of cortical networks are under genetic control. Using twin and family data from the Human Connectome Project (n=1,023), we find increased variability and reduced heritability in the size of heteromodal association networks (h2: M=0.33, SD=0.071), relative to unimodal sensory/motor cortex (h2: M=0.44, SD=0.051). We then demonstrate that the spatial layout of cortical networks is influenced by genetics, using our multi-dimensional estimation of heritability (h2-multi; M=0.14, SD=0.015). However, topographic heritability did not differ between heteromodal and unimodal networks. Genetic factors had a regionally variable influence on brain organization, such that the heritability of network topography was greatest in prefrontal, precuneus, and posterior parietal cortex. Taken together, these data are consistent with relaxed genetic control of association cortices relative to primary sensory/motor regions, and have implications for understanding population-level variability in brain functioning, guiding both individualized prediction and the interpretation of analyses that integrate genetics and neuroimaging.

show abstract

“…refs. [32][33][34][35] ). However, the majority of previously used fMRIbased predictive models were only internally validated (i.e.…”

Section: Discussionmentioning

confidence: 99%

Pain-free resting-state functional brain connectivity predicts individual pain sensitivity

Spisák¹,

Kincses

Schlitt³

et al. 2020

Nat Commun

View full text Add to dashboard Cite

Individual differences in pain perception are of interest in basic and clinical research as altered pain sensitivity is both a characteristic and a risk factor for many pain conditions. It is, however, unclear how individual sensitivity to pain is reflected in the pain-free resting-state brain activity and functional connectivity. Here, we identify and validate a network pattern in the pain-free resting-state functional brain connectome that is predictive of interindividual differences in pain sensitivity. Our predictive network signature allows assessing the individual sensitivity to pain without applying any painful stimulation, as might be valuable in patients where reliable behavioural pain reports cannot be obtained. Additionally, as a direct, non-invasive readout of the supraspinal neural contribution to pain sensitivity, it may have implications for translational research and the development and assessment of analgesic treatment strategies.

show abstract

Deep Neural Networks and Kernel Regression Achieve Comparable Accuracies for Functional Connectivity Prediction of Behavior and Demographics

Cited by 68 publications

References 112 publications

Predicting Alzheimer’s disease progression using deep recurrent neural networks

Predicting Alzheimer’s disease progression using deep recurrent neural networks

Heritability of individualized cortical network topography

Pain-free resting-state functional brain connectivity predicts individual pain sensitivity

Contact Info

Product

Resources

About