Ensemble learning with 3D convolutional neural networks for connectome-based prediction
This work addresses the variability in fMRI analysis for neuroscience and medical diagnosis, though it is incremental as it builds on existing ensemble and CNN methods.
The study tackled the problem of sensitivity to brain parcellation choices in resting-state fMRI machine learning models by proposing an ensemble learning strategy with a novel 3D CNN approach, which performed as well as models using standard atlases and showed promising results in autism classification and age prediction tasks.
The specificty and sensitivity of resting state functional MRI (rs-fMRI) measurements depend on pre-processing choices, such as the parcellation scheme used to define regions of interest (ROIs). In this study, we critically evaluate the effect of brain parcellations on machine learning models applied to rs-fMRI data. Our experiments reveal a remarkable trend: On average, models with stochastic parcellations consistently perform as well as models with widely used atlases at the same spatial scale. We thus propose an ensemble learning strategy to combine the predictions from models trained on connectivity data extracted using different (e.g., stochastic) parcellations. We further present an implementation of our ensemble learning strategy with a novel 3D Convolutional Neural Network (CNN) approach. The proposed CNN approach takes advantage of the full-resolution 3D spatial structure of rs-fMRI data and fits non-linear predictive models. Our ensemble CNN framework overcomes the limitations of traditional machine learning models for connectomes that often rely on region-based summary statistics and/or linear models. We showcase our approach on a classification (autism patients versus healthy controls) and a regression problem (prediction of subject's age), and report promising results.