Regression-based machine-learning approaches to predict task activation using resting-state fMRI

Alexander D. Cohen, Ziyi Chen, Oiwi Parker Jones, Chen Niu, Yang Wang

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Resting-state fMRI has shown the ability to predict task activation on an individual basis by using a general linear model (GLM) to map resting-state network features to activation z-scores. The question remains whether the relatively simplistic GLM is the best approach to accomplish this prediction. In this study, several regression-based machine-learning approaches were compared, including GLMs, feed-forward neural networks, and random forest bootstrap aggregation (bagging). Resting-state and task data from 350 Human Connectome Project subjects were analyzed. First, the effect of the number of training subjects on the prediction accuracy was evaluated. In addition, the prediction accuracy and Dice coefficient were compared across models. Prediction accuracy increased with the training number up to 200 subjects; however, an elbow in the prediction curve occurred around 30–40 training subjects. All models performed well with correlation matrices, which displayed correlation between actual and predicted task activation for all subjects, exhibiting a strong diagonal trend for all tasks. Overall, the neural network and random forest bagging techniques outperformed the GLM. These approaches, however, require additional computing power and processing time. These results show that, while the GLM performs well, resting-state fMRI prediction of task activation could benefit from more complex machine learning approaches.

Original languageEnglish (US)
JournalHuman Brain Mapping
DOIs
StateAccepted/In press - Jan 1 2019
Externally publishedYes

Fingerprint

Linear Models
Magnetic Resonance Imaging
Connectome
Elbow
Machine Learning
Forests

Keywords

  • fMRI
  • machine learning
  • neural networks
  • random-forest bootstrap aggregation
  • resting state

ASJC Scopus subject areas

  • Anatomy
  • Radiological and Ultrasound Technology
  • Radiology Nuclear Medicine and imaging
  • Neurology
  • Clinical Neurology

Cite this

Regression-based machine-learning approaches to predict task activation using resting-state fMRI. / Cohen, Alexander D.; Chen, Ziyi; Parker Jones, Oiwi; Niu, Chen; Wang, Yang.

In: Human Brain Mapping, 01.01.2019.

Research output: Contribution to journalArticle

@article{4ab68bef4b3743c1adfe98bb202b5140,
title = "Regression-based machine-learning approaches to predict task activation using resting-state fMRI",
abstract = "Resting-state fMRI has shown the ability to predict task activation on an individual basis by using a general linear model (GLM) to map resting-state network features to activation z-scores. The question remains whether the relatively simplistic GLM is the best approach to accomplish this prediction. In this study, several regression-based machine-learning approaches were compared, including GLMs, feed-forward neural networks, and random forest bootstrap aggregation (bagging). Resting-state and task data from 350 Human Connectome Project subjects were analyzed. First, the effect of the number of training subjects on the prediction accuracy was evaluated. In addition, the prediction accuracy and Dice coefficient were compared across models. Prediction accuracy increased with the training number up to 200 subjects; however, an elbow in the prediction curve occurred around 30–40 training subjects. All models performed well with correlation matrices, which displayed correlation between actual and predicted task activation for all subjects, exhibiting a strong diagonal trend for all tasks. Overall, the neural network and random forest bagging techniques outperformed the GLM. These approaches, however, require additional computing power and processing time. These results show that, while the GLM performs well, resting-state fMRI prediction of task activation could benefit from more complex machine learning approaches.",
keywords = "fMRI, machine learning, neural networks, random-forest bootstrap aggregation, resting state",
author = "Cohen, {Alexander D.} and Ziyi Chen and {Parker Jones}, Oiwi and Chen Niu and Yang Wang",
year = "2019",
month = "1",
day = "1",
doi = "10.1002/hbm.24841",
language = "English (US)",
journal = "Human Brain Mapping",
issn = "1065-9471",
publisher = "Wiley-Liss Inc.",

}

TY - JOUR

T1 - Regression-based machine-learning approaches to predict task activation using resting-state fMRI

AU - Cohen, Alexander D.

AU - Chen, Ziyi

AU - Parker Jones, Oiwi

AU - Niu, Chen

AU - Wang, Yang

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Resting-state fMRI has shown the ability to predict task activation on an individual basis by using a general linear model (GLM) to map resting-state network features to activation z-scores. The question remains whether the relatively simplistic GLM is the best approach to accomplish this prediction. In this study, several regression-based machine-learning approaches were compared, including GLMs, feed-forward neural networks, and random forest bootstrap aggregation (bagging). Resting-state and task data from 350 Human Connectome Project subjects were analyzed. First, the effect of the number of training subjects on the prediction accuracy was evaluated. In addition, the prediction accuracy and Dice coefficient were compared across models. Prediction accuracy increased with the training number up to 200 subjects; however, an elbow in the prediction curve occurred around 30–40 training subjects. All models performed well with correlation matrices, which displayed correlation between actual and predicted task activation for all subjects, exhibiting a strong diagonal trend for all tasks. Overall, the neural network and random forest bagging techniques outperformed the GLM. These approaches, however, require additional computing power and processing time. These results show that, while the GLM performs well, resting-state fMRI prediction of task activation could benefit from more complex machine learning approaches.

AB - Resting-state fMRI has shown the ability to predict task activation on an individual basis by using a general linear model (GLM) to map resting-state network features to activation z-scores. The question remains whether the relatively simplistic GLM is the best approach to accomplish this prediction. In this study, several regression-based machine-learning approaches were compared, including GLMs, feed-forward neural networks, and random forest bootstrap aggregation (bagging). Resting-state and task data from 350 Human Connectome Project subjects were analyzed. First, the effect of the number of training subjects on the prediction accuracy was evaluated. In addition, the prediction accuracy and Dice coefficient were compared across models. Prediction accuracy increased with the training number up to 200 subjects; however, an elbow in the prediction curve occurred around 30–40 training subjects. All models performed well with correlation matrices, which displayed correlation between actual and predicted task activation for all subjects, exhibiting a strong diagonal trend for all tasks. Overall, the neural network and random forest bagging techniques outperformed the GLM. These approaches, however, require additional computing power and processing time. These results show that, while the GLM performs well, resting-state fMRI prediction of task activation could benefit from more complex machine learning approaches.

KW - fMRI

KW - machine learning

KW - neural networks

KW - random-forest bootstrap aggregation

KW - resting state

UR - http://www.scopus.com/inward/record.url?scp=85076511744&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85076511744&partnerID=8YFLogxK

U2 - 10.1002/hbm.24841

DO - 10.1002/hbm.24841

M3 - Article

C2 - 31638304

AN - SCOPUS:85076511744

JO - Human Brain Mapping

JF - Human Brain Mapping

SN - 1065-9471

ER -