Ensemble learning with 3D convolutional neural networks for functional connectome-based prediction
| Autor: | Keith Jamison, Amy Kuceyeski, Meenakshi Khosla, Mert R. Sabuncu |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Adult
Male Scheme (programming language) Adolescent Autism Spectrum Disorder Computer science Cognitive Neuroscience Machine learning computer.software_genre Convolutional neural network Article 050105 experimental psychology Cohort Studies Machine Learning Young Adult 03 medical and health sciences Atlases as Topic 0302 clinical medicine Image Interpretation Computer-Assisted Connectome medicine Humans 0501 psychology and cognitive sciences Sensitivity (control systems) Child computer.programming_language Resting state fMRI business.industry 05 social sciences Linear model Brain medicine.disease Magnetic Resonance Imaging Ensemble learning Neurology Autism spectrum disorder Autism Neural Networks Computer Artificial intelligence business computer 030217 neurology & neurosurgery |
| Zdroj: | Neuroimage |
| ISSN: | 1053-8119 |
| Popis: | The specificity and sensitivity of resting state functional MRI (rs-fMRI) measurements depend on preprocessing 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 an intriguing 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. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect