Spatially guided functional correlation tensor: A new method to associate body mass index and white matter neuroimaging

https://doi.org/10.1016/j.compbiomed.2019.02.010Get rights and content

Highlights

  • We developed an enhanced functional correlation tensor using spatial information.

  • The method was applied to mimic real diffusion tensor imaging using functional MRI.

  • We used the method to find imaging biomarkers for obesity.

Abstract

Obesity causes critical health problems including cardiovascular disease, diabetes, and stroke. Various neuroimaging methods including diffusion tensor imaging (DTI) are used to explore white matter (WM) alterations in obesity. The functional correlation tensor (FCT) is a method to simulate DTI in WM using resting-state functional magnetic resonance imaging (rs-fMRI). In this study, we enhanced the FCT with additional anatomical information from T1-weighted data in a regression framework. The goal was to 1) develop a spatially guided enhanced FCT (s-eFCT) and to 2) use it to identify imaging biomarkers for obesity. We computed fractional anisotropy (FA) and the mean diffusivity (MD) from the s-eFCT. The regional FA and MD values that can explain body mass index (BMI) well were chosen. The identified regional FA and MD values were used to predict BMI values. The correlation between real and predicted BMIs was 0.57. There was no significant correlation between real and predicted DTI using the MD. The BMI predicted using FA was used to classify participants into three obesity subgroups. The classification accuracy was 57.20%. In summary, we found potential imaging biomarkers of obesity based on the s-eFCT.

Introduction

Obesity is a worldwide health problem, which is linked to many diseases including cardiovascular disease, diabetes, and stroke [1]. Body mass index (BMI) is a simple measure of obesity [2], and it is a representative measurement to assess the accumulation of body fat [3]. Previous studies found a positive correlation between BMI and morbidity suggesting that BMI might be an indicator for the physical condition of people with obesity [2,3].

Many studies found that obesity is associated with brain based on neuroimaging results of magnetic resonance imaging (MRI), positron emission tomography, and single-photon emission computed tomography (SPECT) [[4], [5], [6], [7], [8], [9]]. MRI is an especially useful tool as it provides both structural and functional information of the brain. Diffusion tensor imaging (DTI) is one of the MRI modalities that measures the degree of water diffusion in white matter (WM). The directionality of diffusion is measured using fractional anisotropy (FA), and the magnitude of diffusion is measured using the mean diffusivity (MD). FA and the MD are the representative scalar measurements computed from DTI, and many studies adopt them to quantify the WM properties [10]. Previous neuroimaging studies found significant relationships between the WM structure alterations and obesity using DTI [[11], [12], [13]]. Karlsson et al. found that WM atrophy was associated with the percentage of body fat [11]. Bolzenius et al. adopted tractography analysis and found a negative correlation between BMI and the length of fibers in the temporal lobe [12]. Stanek et al. found that an increased BMI was related to decreased FA in the corpus callosum and fornix [13]. In a previous study, we predicted BMI using the structural connectivity of DTI [14]. These studies collectively suggest that obesity is related to the altered structure of WM.

Functional MRI (fMRI) that measures the blood-oxygen-level dependent (BOLD) signal from gray matter (GM) is a widely adopted neuroimaging modality in obesity-related studies [15,16]. Existing fMRI studies focused on analyzing the BOLD signal from GM [[17], [18], [19], [20]]. Still, a few studies reported that significant BOLD signal existed in WM in smaller magnitude [21]. Especially, the robust signal was observed in the corpus callosum and internal capsule, parts of WM [[21], [22], [23], [24]]. These studies collectively provided a basis for exploring the BOLD signal in the WM. A method to explore BOLD signals from WM using fMRI was proposed by Ding et al. [25]. They developed a method known as the functional correlation tensor (FCT) to mimic DTI [25,26]. The FCT was constructed by calculating the correlations of time courses between a given voxel and its adjacent voxels, and then the correlation values were further modified with a dyadic tensor. Ding et al. reconstructed the neuronal fibers in the corpus callosum and optic radiation using FCT suggesting the possibility of developing pseudo-DTI from fMRI [25]. This might allow us to extract DTI like information from fMRI data and thus might save the scanning time as there is less need to perform real DTI data acquisition. The decrease in the scanning time can be an important factor for children and elderly subjects who might have difficulties with staying in the scanner for an extended period of time, which is required for multimodal acquisition studies.

In this study, we aimed to achieve two goals. First, we aimed to enhance the FCT by incorporating spatial information of T1-weighted structural data. We hypothesized that the spatial information might provide complementary information for constructing the FCT. Second, we aimed to find biomarkers of obesity using a spatially guided enhanced FCT (s-eFCT). We predicted BMI using the s-eFCT. We hypothesized that the s-eFCT might be used as imaging biomarkers to distinguish among subclasses of obesity.

Section snippets

Subjects and imaging data

The institutional review broad (IRB) of Sungkyunkwan University approved our study. Consent was waived for this retrospective study. Our study was performed in full accordance with local IRB guidelines. We obtained DTI, resting-state functional magnetic resonance imaging (rs-fMRI), and T1-weighted data from enhanced Nathan Kline Institute-Rockland sample (NKI-RS) database Release 1–8 [27]. All NKI-RS participants were scanned on a Siemens Magnetom Trio Tim scanner with the following imaging

Study participants

We considered 264 subjects in this study. The subjects were classified into three subgroups of HW, OW and OB groups. Detailed demographic information is described in Table 1 for all three subgroups.

Quality of the s-eFCT

We constructed the s-eFCT using the fMRI with T1-weighted data and calculated the FA and MD values for the WM regions. The tensors were compared with those of the real diffusion tensor and the s-eFCT using Pearson's correlation in the 10-fold cross-validation. The mean correlation of the tensors over

Discussion

In the current study, we aimed to develop a new variant of the FCT, the s-eFCT, which contains both functional and structural brain information using fMRI and T1-weighted data. The functional information was measured by correlating temporal fluctuations in fMRI between all pair of voxels in the WM. The structural information was voxel-wise T1-weighted data. We used our s-eFCT to predict the BMIs of people with a wide range of BMI values, and it led to fair prediction results (mean r = 0.57).

Conclusions

In this study, we combined the FCT and T1-weighted anatomical information to construct our s-eFCT in a regression framework to model the diffusion tensor. The s-eFCT model was used to compute the FA and MD values of the WM regions. We identified FA and MD of the WM regions related to BMI, which were used as input of a BMI score prediction model. The correlation between the real and predicted BMIs was 0.57. We divided the predicted BMI to distinguish between three subgroups of obesity. The

Acknowledgments

This study was supported by the Institute for Basic Science (grant number IBS-R015-D1), the National Research Foundation of Korea (grant numbers NRF-2016R1A2B4008545 and NRF-2016H1A2A1907833), and the Ministry of Science and ICT of Korea under the ITRC program (IITP-2018-2018-0-01798). Imaging data were obtained from the NKI-RS database.

References (46)

  • Z. Ding et al.

    Visualizing functional pathways in the human brain using correlation tensors and magnetic resonance imaging

    Magn. Reson. Imaging

    (2016)
  • M. Jenkinson et al.

    Fsl, Neuroimage.

    (2012)
  • R.W. Cox

    AFNI: software for analysis and visualization of functional magnetic resonance neuroimages

    Comput. Biomed. Res.

    (1996)
  • A.J. Schwarz et al.

    Negative edges and soft thresholding in complex network analysis of resting state functional connectivity data

    Neuroimage

    (2011)
  • Y. Zhou et al.

    Functional MRI registration with tissue-specific patch-based functional correlation tensors

    Hum. Brain Mapp.

    (2018)
  • S. Mori et al.

    Stereotaxic white matter atlas based on diffusion tensor imaging in an ICBM template

    Neuroimage

    (2008)
  • I. Papageorgiou et al.

    Abnormalities of brain neural circuits related to obesity: a Diffusion Tensor Imaging study

    Magn. Reson. Imaging

    (2017)
  • L. Zhang et al.

    Learning-based structurally-guided construction of resting-state functional correlation tensors

    Magn. Reson. Imaging

    (2017)
  • F.Q. Nuttall

    Body mass index: obesity, BMI, and health: a critical review

    Nutr. Today

    (2015)
  • K.E. Bradbury et al.

    Association between physical activity and body fat percentage, with adjustment for BMI: a large cross-sectional analysis of UK Biobank

    Br. Med. J. Open

    (2017)
  • K.C. Willeumier et al.

    Elevated BMI is associated with decreased blood flow in the prefrontal cortex using SPECT imaging in healthy adults

    Obesity

    (2011)
  • A. Del Parigi et al.

    Neuroimaging and obesity: mapping the brain responses to hunger and satiation in humans using positron emission tomography

    Ann. N. Y. Acad. Sci.

    (2002)
  • Y. Assaf et al.

    Diffusion tensor imaging (DTI)-based white matter mapping in brain research: a review

    J. Mol. Neurosci.

    (2008)
  • Cited by (8)

    View all citing articles on Scopus
    1

    These authors contributed equally to this work.

    View full text