Refining visceral adipose tissue quantification: Influence of sex, age, and BMI on single slice estimation in 3D MRI of the German National Cohort

Tobias Haueise, Fritz Schick, Norbert Stefan, Elena Grune, Marc-Nicolas von Itter, Hans‐Ulrich Kauczor, Johanna Nattenmüller, Tobias Norajitra, Tobias Nonnenmacher, Susanne Rospleszcz, Klaus Maier‐Hein, Christopher L. Schlett, Jakob Weiss, Beate Fischer, Karl‐Heinz Jöckel, Lilian Krist, Thoralf Niendorf, Annette Peters, Anja M. Sedlmeier, Stefan N. Willich, Fabian Bamberg, Jürgen MachannZeitschrift 2025 · DOI

TL;DR

This study refines how we estimate visceral adipose tissue (VAT) volume from single MRI slices by accounting for sex, age, and BMI. They found that the optimal slice location for accurate VAT estimation varies significantly based on these factors, particularly for individuals with obesity, offering a more personalized approach to this crucial health metric.

What Problem Does This Paper Address?

Visceral adipose tissue (VAT) is a key indicator of metabolic syndrome and associated health risks like type 2 diabetes and cardiovascular disease. While whole-body MRI provides a gold standard for quantifying VAT, it's resource-intensive. Single-slice MRI offers a more efficient alternative, but its accuracy is hampered by not accounting for how VAT distribution changes across the body. Previous studies have identified potential reference slices (like L3), but they often overlooked the significant influence of individual characteristics such as sex, age, and Body Mass Index (BMI) on VAT distribution. This paper aims to bridge this gap by systematically analyzing these influences to provide more accurate, personalized single-slice VAT estimation methods.

Methods

The researchers utilized a secondary analysis of 3D VIBE Dixon MRI data from the German National Cohort (NAKO), including over 11,000 participants. They employed a deep learning-based U-Net model to automatically segment visceral adipose tissue (VAT) from the trunk, from the femoral heads to the cardiac apex. To standardize VAT distribution analysis, they derived craniocaudal VAT profiles by normalizing per-slice VAT percentages to anatomical landmarks like vertebral bodies (L5 to Th10) and intervertebral discs.

A key innovation was the development of a method to standardize locations caudally to L5, down to the femoral heads, using linear interpolation of VAT pixel data. This allowed for the creation of a scale with a 0.5 cm step size. To identify age-independent reference locations for single-slice VAT estimation, they analyzed these craniocaudal profiles within specific BMI and sex groups, looking for points of intersection where age-related differences were minimized. The final VAT volume estimation from a single slice was then calculated using a sex- and BMI-specific scaling factor derived from the cohort data.

Datasets

The study used data from the German National Cohort (NAKO). This prospective, population-based cohort study includes over 200,000 participants, with a subset of approximately 30,000 undergoing whole-body MRI. The specific data used in this study comprised 3D VIBE Dixon MRI from 11,191 participants screened between May 2014 and December 2016. After applying exclusion criteria (imaging errors, missing measurements, underweight participants, and small age groups), the final analysis included 11,036 participants.

Key Insights and Ideas

  • VAT Distribution is Highly Personalized: The study strongly emphasizes that VAT distribution isn't uniform and is significantly influenced by sex, age, and BMI. This challenges the idea of a single "best" slice for everyone.
  • BMI Drives Caudal Shifts: As BMI increases, the optimal reference slice for VAT estimation tends to shift towards more caudal (lower) locations. This suggests that individuals with higher BMI have a different distribution of visceral fat.
  • Obesity in Women Requires Special Attention: For women with obesity (BMI >30 kg/m²), the L1 vertebral level emerged as a particularly optimal reference location, highlighting a specific need for tailored approaches in this demographic.
  • Age Independence is Achievable (with caveats): By identifying specific "intersection points" in the craniocaudal VAT profiles, the researchers could minimize age-related variability, enabling more reliable age-independent single-slice estimations.
  • Deep Learning for Standardization: The use of automated deep learning segmentation and spine co-registration was crucial for standardizing anatomical landmarks across a large cohort, overcoming manual segmentation challenges.

Key Results

  • The final study population comprised 11,036 participants (mean age 52 ± 11 years, 5681 men).
  • Craniocaudal VAT distribution showed qualitative differences between men and women, and varied with age and BMI.
  • Age-independent single-slice VAT estimates demonstrated strong correlations with reference VAT volumes (Pearson's r ranging from 0.77-0.88 in men and 0.72-0.89 in women).
  • Anatomical locations for accurate VAT estimation varied significantly with sex and BMI. For instance, intersections were shifted cranially for women with BMI >30 kg/m².
  • Mean errors for single-slice VAT estimation (compared to volumetric quantification) were low: -0.07 ± 0.99 L in men and -0.06 ± 0.56 L in women, corresponding to relative mean absolute errors of 1.44% and 2.37%, respectively.
  • Table 2 provides specific locations (e.g., L5/4, L5-1 cm, L2/1) and scaling factors for VAT estimation based on BMI groups and sex.

Limitations and Future Work

  • Standardization Bias: The reliance on automatic spine segmentation and linear interpolation for landmark standardization is acknowledged as a potential source of bias.
  • Caudal Landmark Limitations: Standardized landmarks located caudally to L5 might have limited practical applicability in some cases.
  • BMI Threshold Ambiguity: The use of discrete BMI thresholds can lead to ambiguities, particularly for individuals near the boundaries (e.g., women around 30 kg/m²).
  • Cross-sectional Data: The study is based on cross-sectional data, limiting the ability to definitively answer whether single-slice measurements can reliably quantify changes in VAT after lifestyle interventions.
  • Cohort Specificity: The findings are primarily based on a cohort where over 95% of participants identified as White, potentially limiting generalizability to other ethnic groups.
  • Disease-Specific VAT: The study did not investigate whether VAT distribution differs in patients with specific metabolic diseases or other conditions that affect weight and fat accumulation.

Future work could explore the correlation of these single-slice VAT estimates with insulin sensitivity and cardiometabolic risk, investigate the impact of these findings in different ethnic groups, and assess their utility in longitudinal studies for tracking changes in VAT over time.

Summary updated

Abstract

The selection of reference locations should be different depending on BMI groups, with a preference for caudal shifts in location with increasing BMI. For women with obesity (BMI >30 kg/m<sup>2</sup>), the L1 level emerges as the optimal reference location.
BibTeX
@inproceedings{haueise2025refining,
  author = {Tobias Haueise and Fritz Schick and Norbert Stefan and Elena Grune and Marc-Nicolas von Itter and Hans‐Ulrich Kauczor and Johanna Nattenmüller and Tobias Norajitra and Tobias Nonnenmacher and Susanne Rospleszcz and Klaus Maier‐Hein and Christopher L. Schlett and Jakob Weiss and Beate Fischer and Karl‐Heinz Jöckel and Lilian Krist and Thoralf Niendorf and Annette Peters and Anja M. Sedlmeier and Stefan N. Willich and Fabian Bamberg and Jürgen Machann},
  title = {{Refining visceral adipose tissue quantification: Influence of sex, age, and BMI on single slice estimation in 3D MRI of the German National Cohort}},
  booktitle = {Zeitschrift},
  year = {2025},
  doi = {10.1016/j.zemedi.2025.02.005},
}
Figure