There is significant interest in the development and application of deep neural networks (DNNs) to neuroimaging data. A growing literature suggests that DNNs outperform their classical counterparts in a variety of neuroimaging applications, yet there are few direct comparisons of relative utility. Here, we compared the performance of three DNN architectures and a classical machine learning algorithm (kernel regression) in predicting individual phenotypes from whole-brain resting-state functional connectivity (RSFC) patterns. One of the DNNs was a generic fully-connected feedforward neural network, while the other two DNNs were recently published approaches specifically designed to exploit the structure of connectome data. By using a combined sample of almost 10,000 participants from the Human Connectome Project (HCP) and UK Biobank, we showed that the three DNNs and kernel regression achieved similar performance across a wide range of behavioral and demographic measures. Furthermore, the generic feedforward neural network exhibited similar performance to the two state-of-the-art connectome-specific DNNs. When predicting fluid intelligence in the UK Biobank, performance of all algorithms dramatically improved when sample size increased from 100 to 1000 subjects. Improvement was smaller, but still significant, when sample size increased from 1000 to 5000 subjects. Importantly, kernel regression was competitive across all sample sizes. Overall, our study suggests that kernel regression is as effective as DNNs for RSFC-based behavioral prediction, while incurring significantly lower computational costs. Therefore, kernel regression might serve as a useful baseline algorithm for future studies.

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.

There is significant interest in using brain imaging data to predict non-brain-imaging phenotypes in individual participants. However, most prediction studies are underpowered, relying on less than a few hundred participants, leading to low reliability and inflated prediction performance. Yet, small sample sizes are unavoidable when studying clinical populations or addressing focused neuroscience questions. Here, we propose a simple framework - "meta-matching" - to translate predictive models from large-scale datasets to new unseen non-brain-imaging phenotypes in boutique studies. The key observation is that many large-scale datasets collect a wide range inter-correlated phenotypic measures. Therefore, a unique phenotype from a boutique study likely correlates with (but is not the same as) some phenotypes in some large-scale datasets. Meta-matching exploits these correlations to boost prediction in the boutique study. We applied meta-matching to the problem of predicting non-brain-imaging phenotypes using resting-state functional connectivity (RSFC). Using the UK Biobank (N = 36,848), we demonstrated that meta-matching can boost the prediction of new phenotypes in small independent datasets by 100% to 400% in many scenarios. When considering relative prediction performance, meta-matching significantly improved phenotypic prediction even in samples with 10 participants. When considering absolute prediction performance, meta-matching significantly improved phenotypic prediction when there were least 50 participants. With a growing number of large-scale population-level datasets collecting an increasing number of phenotypic measures, our results represent a lower bound on the potential of meta-matching to elevate small-scale boutique studies.

The results of most neuroimaging studies are reported in volumetric (e.g., MNI152) or surface (e.g., fsaverage) coordinate systems. Accurate mappings between volumetric and surface coordinate systems can facilitate many applications, such as projecting fMRI group analyses from MNI152/Colin27 to fsaverage for visualization, or projecting resting-state fMRI parcellations from fsaverage to MNI152/Colin27 for volumetric analysis of new data. However, there has been surprisingly little research on this topic. Here, we evaluated three approaches for mapping data between MNI152/Colin27 and fsaverage coordinate systems by simulating the above applications: projection of group-average data from MNI152/Colin27 to fsaverage and projection of fsaverage parcellations to MNI152/Colin27. Two of the approaches are currently widely used. A third approach (registration fusion) was previously proposed, but not widely adopted. Two implementations of the registration fusion (RF) approach were considered, with one implementation utilizing the Advanced Normalization Tools (ANTs). We found that RF-ANTs performed the best for mapping between fsaverage and MNI152/Colin27, even for new subjects registered to MNI152/Colin27 using a different software tool (FSL FNIRT). This suggests that RF-ANTs would be useful even for researchers not using ANTs. Finally, it is worth emphasizing that the most optimal approach for mapping data to a coordinate system (e.g., fsaverage) is to register individual subjects directly to the coordinate system, rather than via another coordinate system. Only in scenarios where the optimal approach is not possible (e.g., mapping previously published results from MNI152 to fsaverage), should the approaches evaluated in this manuscript be considered. In these scenarios, we recommend RF-ANTs (https://github.com/ThomasYeoLab/CBIG/tree/master/stable_projects/registration/Wu 2017_RegistrationFusion).