Reconstructing lost BOLD signal in individual participants using deep machine learning

Yuxiang Yan,Xiaolong Peng,Louisa Dahmani,Shijun Li,Jianxun Ren,Lunhao Shen,Jianguo Zhang,Changqing Jiang,Ruiqi Wang,Meiyun Wang,Hesheng Liu,Luming Li,Changgeng He,Chen Gong,Ye Tian,Yuanxiang Lin,Yi Guo,Bo Hong

doi:10.1038/s41467-020-18823-9

Abstract

Signal loss in blood oxygen level-dependent (BOLD) functional neuroimaging is common and can lead to misinterpretation of findings. Here, we reconstructed compromised fMRI signal using deep machine learning. We trained a model to learn principles governing BOLD activity in one dataset and reconstruct artificially compromised regions in an independent dataset, frame by frame. Intriguingly, BOLD time series extracted from reconstructed frames are correlated with the original time series, even though the frames do not independently carry any temporal information. Moreover, reconstructed functional connectivity maps exhibit good correspondence with the original connectivity maps, indicating that the model recovers functional relationships among brain regions. We replicated this result in two healthy datasets and in patients whose scans suffered signal loss due to intracortical electrodes. Critically, the reconstructions capture individual-specific information. Deep machine learning thus presents a unique opportunity to reconstruct compromised BOLD signal while capturing features of an individual’s own functional brain organization.

Highlights

Signal loss in blood oxygen level-dependent (BOLD) functional neuroimaging is common and can lead to misinterpretation of findings
We reconstructed BOLD signals in three cohorts: healthy young adults (GSP and Human Connectome Project (HCP) data sets) with artificially compromised cortical regions and patients whose functional magnetic resonance imaging (fMRI) scans suffered from interference due to metal implants
Our results indicate that such a machine learning technique successfully reconstructs individual-specific BOLD signals and can approximate the functional connectivity patterns observed in the intact or unimpaired state

Summary

Introduction

Signal loss in blood oxygen level-dependent (BOLD) functional neuroimaging is common and can lead to misinterpretation of findings. Reconstructed functional connectivity maps exhibit good correspondence with the original connectivity maps, indicating that the model recovers functional relationships among brain regions We replicated this result in two healthy datasets and in patients whose scans suffered signal loss due to intracortical electrodes. Intracortical electrodes implanted in patients interfere with the BOLD signal, potentially due to their lead connectors, resulting in a significant signal loss in brain regions close to the connection site on the skull[1]. This hampers studies that investigate whole-brain activity and functional connectivity and may result in misinterpretation of findings. We hypothesized that the reconstructed BOLD signals from the post-surgical data would be highly similar to the pre-surgical BOLD signals and that they would reflect patterns of activity that are specific to the individual

Objectives

Methods

Conclusion