To BMI, or not to BMI? Adiposity estimation from head T1-weighted MR images
Poster No:
2102
Submission Type:
Abstract Submission
Authors:
Robert Dahnke1, Polona Kalc2, Felix Hoffstaedter3, Eileen Luders4, Christian Gaser1
Institutions:
1Jena University Hospital, Jena, Germany, 2University Hospital Jena, Jena, Germany, 3Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich, Jülich, NRW, 4University of Auckland, Auckland, New Zealand
First Author:
Co-Author(s):
Felix Hoffstaedter
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Jülich, NRW
Institute of Neuroscience and Medicine (INM-7: Brain and Behaviour), Research Centre Jülich
Jülich, NRW
Introduction:
Obesity has adverse effects on cognitive and brain health [1–2]. Nevertheless, most neuroimaging studies do not collect sophisticated body composition measures and usually rely on simple approximations by calculating the body mass index (BMI) from weight and height data. This approach has several limitations, especially when used in cohorts of older individuals [3–4]. To overcome this drawback, we present a novel method to approximate body adiposity by measuring the thickness of the head using T1-weighted brain images.
Methods:
Sample
This research has been conducted using a subsample of 1000 healthy subjects (M = 63.81±6.27 years, 50% women) from the UK Biobank (UKB application 41655). Imaging parameters are provided in [5]. The sample was used to extract and validate our adiposity estimate against the body composition measures, obtained by abdominal MRI [6] and processed by AMRA Profiler Research [7]. In addition, a subsample of 63 subjects from OASIS-3 (M = 69.27±7.32 years, 54% women) was used to determine the retest reliability. Imaging parameters are provided in [8].
Data processing
All T1-weighted brain images were processed using SPM12 (https://www.fil.ion.ucl.ac.uk/spm/) and Matlab 2021a (https://www.mathworks.com), which produced the following segments: gray matter, white matter, cerebrospinal fluid, skull, soft head tissue, and background. Of note, instead of the default setting (3 mm), we set SPM's 'samp' parameter to 5 mm to ensure that non-brain tissues were properly classified.
To approximate the adipose tissue of the head, we estimated the local thickness of the segment classified as soft head tissue. This was achieved by calculating the shortest distance from each soft head tissue voxel to the skull and to the background. The sum of both distances then yielded the estimate of the voxel-wise head thickness. A separation into different head tissues (i.e., muscle, skin, and fat) was omitted because of varying amounts of (chemical shift) artifacts and inhomogeneities across the image. Of note, the lower portions of all brain scans as well as voxels located more than 30 mm from the skull were excluded from the head thickness estimation to avoid side effects due to defacing and/or varying scanning protocols and procedures.
This research has been conducted using a subsample of 1000 healthy subjects (M = 63.81±6.27 years, 50% women) from the UK Biobank (UKB application 41655). Imaging parameters are provided in [5]. The sample was used to extract and validate our adiposity estimate against the body composition measures, obtained by abdominal MRI [6] and processed by AMRA Profiler Research [7]. In addition, a subsample of 63 subjects from OASIS-3 (M = 69.27±7.32 years, 54% women) was used to determine the retest reliability. Imaging parameters are provided in [8].
Data processing
All T1-weighted brain images were processed using SPM12 (https://www.fil.ion.ucl.ac.uk/spm/) and Matlab 2021a (https://www.mathworks.com), which produced the following segments: gray matter, white matter, cerebrospinal fluid, skull, soft head tissue, and background. Of note, instead of the default setting (3 mm), we set SPM's 'samp' parameter to 5 mm to ensure that non-brain tissues were properly classified.
To approximate the adipose tissue of the head, we estimated the local thickness of the segment classified as soft head tissue. This was achieved by calculating the shortest distance from each soft head tissue voxel to the skull and to the background. The sum of both distances then yielded the estimate of the voxel-wise head thickness. A separation into different head tissues (i.e., muscle, skin, and fat) was omitted because of varying amounts of (chemical shift) artifacts and inhomogeneities across the image. Of note, the lower portions of all brain scans as well as voxels located more than 30 mm from the skull were excluded from the head thickness estimation to avoid side effects due to defacing and/or varying scanning protocols and procedures.
Results:
The obtained head thickness estimate was validated against BMI (UKB #21001-2.0), body fat percentage (UKB #23099-2.0), abdominal subcutaneous adipose tissue (ASAT) volume (UKB #22408-2.0), visceral adipose tissue volume (VAT; UKB #22407-2.0), and waist circumference (UKB #48-2.0). Our head thickness estimate is moderately associated with BMI in the total sample (r = 0.53, p < 0.001), but more so in separate samples for men and women, respectively (r = 0.63 and r = 0.62, p < 0.001). Moreover, it is highly associated with visceral adipose tissue volume (r = 0.74, p < 0.001). The results of the validation are shown in Figure 2.
In addition, we calculated the retest reliability of the skull BMD estimates using T1-weighted images from the OASIS-3 dataset acquired at two time points within an interval of less than 3 months (r = 0.98; p < 0.001).
In addition, we calculated the retest reliability of the skull BMD estimates using T1-weighted images from the OASIS-3 dataset acquired at two time points within an interval of less than 3 months (r = 0.98; p < 0.001).
Conclusions:
We developed an approximation of subcutaneous fat by making use of a tissue class (i.e., soft head tissue) that is normally discarded when processing brain MR images. Head thickness may provide valuable information beyond the typically used BMI, which has several limitations, especially when used in cohorts of older subjects [3,4]. As open-source brain imaging databases predominantly consist of aging adults, the adiposity measure presents a viable alternative to BMI. Our novel measure may not only find application in basic research as a marker for brain health and aging but also in intervention studies and clinical settings as an indicator (or predictor) for the effectiveness of therapies and interventions.
Modeling and Analysis Methods:
Classification and Predictive Modeling
Image Registration and Computational Anatomy 2
Methods Development
Segmentation and Parcellation
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Anatomy and Functional Systems 1
Keywords:
Morphometrics
STRUCTURAL MRI
Other - obesity
1|2Indicates the priority used for review
Provide references using author date format
[2] Farruggia, M. C., & Small, D. M. (2019), ‘Effects of adiposity and metabolic dysfunction on cognition: A review’, Physiology & Behavior, vol. 208, 112578. https://doi.org/10.1016/j.physbeh.2019.112578
[3] Bhurosy, T., & Jeewon, R. (2013), ‘Pitfalls of Using Body Mass Index (BMI) in Assessment of Obesity Risk’, Current Research in Nutrition and Food Science Journal, vol. 1, no. 1, pp. 71–76. https://doi.org/10.12944/CRNFSJ.1.1.07
[4] Rothman, K. J. (2008), ‘BMI-related errors in the measurement of obesity’, International Journal of Obesity, vol. 32, no. S3, pp. S56–S59. https://doi.org/10.1038/ijo.2008.87
[5] Smith, S.M., Alfaro-Almagro, F., & Miller, K.L. (2020), ‘UK Biobank imaging documentation, Version 1.8.’, biobank.ctsu.ox.ac.uk/crystal/docs/brain_mri.pdf
[6] Linge, J., Borga, M., West, J., Tuthill, T., Miller, M. R., Dumitriu, A., Thomas, E. L., Romu, T., Tunón, P., Bell, J. D., & Dahlqvist Leinhard, O. (2018), ‘Body Composition Profiling in the UK Biobank Imaging Study’, Obesity, vol. 26, no. 11, pp. 1785–1795. https://doi.org/10.1002/oby.22210
[7] Borga, M., Thomas, E. L., Romu, T., Rosander, J., Fitzpatrick, J., Dahlqvist Leinhard, O., & Bell, J. D. (2015), ‘Validation of a fast method for quantification of intra-abdominal and subcutaneous adipose tissue for large-scale human studies: Quantification of IAAT and ASAT’, NMR in Biomedicine, vol. 28, no. 12, pp. 1747–1753. https://doi.org/10.1002/nbm.3432
[8] LaMontagne, P. J., Benzinger, T. Ls., Morris, J. C., Keefe, S., Hornbeck, R., Xiong, C., Grant, E., Hassenstab, J., Moulder, K., Vlassenko, A. G., Raichle, M. E., Cruchaga, C., & Marcus, D. (2019), ‘OASIS-3: Longitudinal Neuroimaging, Clinical, and Cognitive Dataset for Normal Aging and Alzheimer Disease’ [Preprint], Radiology and Imaging. https://doi.org/10.1101/2019.12.13.19014902