Keywords
CT - MR-imaging - physiological studies - treatment effects - CT-quantitative - metabolic Disorders
Background and method
The body composition (BC) of different tissues, particularly, fat, muscle, and bone, can be quantified with radiological methods. Such analyses are particularly important for diseases like overweight and obesity, cachexia (pathological muscle loss), sarcopenia (age-related muscle loss), osteopenia (reduced bone density), and osteoporosis (bone loss with risk of fracture). [Fig. 1] provides a graphical overview of the main BC terms. Computed tomography (CT) and magnetic resonance imaging (MRI), in particular, provide good and nearly anatomical visualization of tissues and accurate information about their spatial distribution. Phenotypes that are associated with a specific disease, e. g., sarcopenic obesity, can be identified with the help of tissue parameters. So far, BC parameters have rarely been considered in daily clinical routine. Radiological findings like “visceral obesity” or “lipomatous changes in the dorsal musculature” are rather uncommon.
Fig. 1 Overview of body composition terms.
This review provides an overview of the current methods and features of a BC analysis. It will present some research highlights and focus on the translation of results into clinical practice. For this purpose, a literature search over the last five years was performed in the American National Library of Medicine (pubmed.gov) for the terms body composition, obesity, sarcopenia, osteopenia in connection with the key words imaging and radiology to identify relevant studies.
The primary aim of this study is to provide a better understanding of the current applications. The focus is on tomographic methods, which are required for a correct spatial visualization of tissues such as visceral adipose tissue (VAT). In principle, BC parameters can also be assessed with simpler methods like DEXA (Dual Energy X-ray Absorptiometry). In practice, however, this will have limitations like the underestimation of visceral adipose tissue (VAT) [1]. A further advantage of tomography is the large selection of available biomarkers. For the musculature, for example, it is possible to determine both the volume as well as the fat infiltration of individual muscles [2]. While conventional BC analysis methods have advantages with respect to availability and feasibility, CT and MRI are more versatile and more reliable. The reader is referred to the literature for further details on the techniques and applications of DEXA and ultrasound imaging [3]
[4]
[5]
[6]
[7]
[8]
[9].
Body composition parameters
Overweight and obesity are defined by the World Health Organization (WHO) as the ratio of body weight to the square of the body size (body mass index) with cutoff values at 25 and 30 kg/m2, respectively [10]. The prevalence of obesity has seen a global increase over the last years. One third of the worldʼs population is currently suffering from overweight or obesity. The term pandemic has been used to describe the situation [11]. Due to its simplicity and feasibility, BMI is still the most widely used measure of BC in spite of the known limitations [12]. Intraindividual changes in BMI during treatment are considered particularly meaningful [13]. Studies on overweight and obesity also rely on radiological methods for BC assessment when anthropometry is not sufficient [14]
[15]
[16], e. g., to differentiate fat from muscle mass [17]. The limited use of anthropometric parameters has recently been highlighted in a longitudinal study with over 3000 participants. Whereas BMI, body weight, or WHR (waist-to-hip-ratio) of the participants remained practically unchanged over the course of two years, significant changes were seen in the visceral and intermuscular fat tissue with MRI [18]. A complete segmentation of all cross-sectional images in the context of a volumetric BC assessment is rather rare. Due to time constraints, the analysis is often restricted to the segmentation of a representative slice, e. g., at the level of the lumbar spine for abdominal fat quantification [1]
[19]
[20]
[21]
[22]. Using refined methods, for example, in longitudinal studies, MRI is also able to detect subtle changes in BC, sometimes even much earlier than conventional methods [18].
In BC analysis of the muscles, both the size and quality can be evaluated, particularly the degree of fatty degeneration. Patients with an oncological primary disease often have a special type of cachexia (cancer cachexia). A loss of skeletal muscle is often accompanied by a functional impairment that may often be partially compensated for, for example, by a special diet [23]. From early on, in an attempt to facilitate the analysis, studies have explored to which degree the loss of muscle mass, e. g., due to sarcopenia, can be detected on a well-defined single slice, often an axial CT scan at the level of the lumbar spine [24]. Similarly, volume and quality of the skeletal muscles can also be quantified by MRI [25].
The BC analysis of bone is traditionally based on the X-ray attenuation measured by DEXA or CT imaging. For peripheral bones, imaging provides its own biomarkers like the diameter, curvature, volume or three-dimensional geometry [26]. For bone tissue, MRI features special techniques to characterize the trabecular bone structure QSM, quantitative susceptibility mapping) and the cortical bone (UTE imaging, ultrashort echo time TE). For the differentiation between osteoporotic and pathological fractures, MRI offers several methods to evaluate the bone marrow fat, especially MR spectroscopy or imaging techniques sensitive to the chemical shift between fat and water signals [27].
The liver is a further target organ for BC analysis, particularly in the context of metabolic dysfunction-associated steatotic liver disease (MASLD). MRI may also be used in that case to determine the fat and water content from the ratio of the respective signals of the hydrogen atoms (PDFF, proton density fat fraction). For that purpose, MRI signals are recorded at more than two points in time (often six) and analyzed by a dedicated software application. MR elastography (MRE) is another functional technique that uses an external, periodic excitation (noninvasive) to generate shear waves in the body, which propagate in a tissue-specific manner. Information about the microscopic tissue displacement during wave propagation is encoded in a series of MR phase images. This data is then processed by a mathematical inversion algorithm to compute an elastogram, which represents the distribution of tissue stiffness. For many years, MRE has been used for the noninvasive evaluation of liver fibrosis, but the technique can also be used in other target regions, e. g., to determine the mechanical and elastic properties of muscular structures [28]
[29]
[30]. Ultrasound-based elastography should also be mentioned here as a BC method that is broadly used for the liver [31] although it is not a tomographic technique. Compared to elastographic methods, MRI diffusion imaging has so far only been used sporadically for BC analysis in spite of its clinical value for many diagnostic questions. There are still some applications, e. g., for the characterization of bone tissue [32]
[33].
In an aging society, the phenotypes of metabolic primary diseases (obesity, osteoporosis, and cachexia) are often observed in combination and will then affect BC. In these cases, bone mineralization and muscle mass will often be reduced, whereas body fat content will be higher [34]. For a standardized BC analysis, age as well as sex and ethnicity must be taken into consideration [35]. Correlating BC parameters with cardiometabolic risk factors, for example, men revealed a pronounced association with intramuscular fat, whereas women showed a (weaker) association for visceral fat instead [36]. Comparing subjects with the same BMI, people of Asian descent often have a higher body fat content, greater abdominal obesity, and a higher intramyocellular lipid and liver fat content than Caucasians [37].
The most promising BC parameters are those that may serve as target variables or biomarkers for novel therapeutic strategies. The physical fitness is often assessed in patients with oncological diseases to evaluate the risks of morbidity and mortality. The disease-related and often unknowing weight loss in cachexia [38] cannot always be differentiated from the age-related physiological loss of muscle mass in sarcopenia [39]
[40]
[41]
[42].
Collaboration between radiology and other diagnostic disciplines including human genetics could help to characterize phenotypes even better [43]. This novel approach to diagnostics is sometimes referred to as integrated diagnostics [44].
Imaging and evaluation methods
Imaging and evaluation methods
Cross-sectional imaging has been used since the 1990s to quantify fat compartments. To date, radiological BC analyses have usually been retrospective and have used, e. g., oncological CT staging data. The usually first task is to segment the corresponding tissue, i. e., to digitally define its margins or contours. [Fig. 2] shows an example of a segmentation of typical fat and muscle groups on a defined axial CT slice. Manual, semiautomatic, and fully automatic analysis tools are available and can be embedded in to the radiology workflow [16]. Some software used for research projects is self-developed and provided for other purposes as well, sometimes as open source code [45]. This allows for a flexible adjustment and extension of functionality, also for third parties, and is usually associated with lower costs. Commercial applications are usually less open but can be used more readily, depending on the degree of certification (e. g., as a medical product), and often have interfaces to the radiological information systems [46]
[47]
[48]. Deep learning methods are used increasingly to automate and accelerate the analysis. Such methods are of particular use for the analysis of large cohorts or on a national level. One prominent example is the management of data collected during the COVID-19 pandemic [49]
[50]
[51].
Fig. 2 Single CT slice at the level of L3 / L4 to evaluate body composition in a 60-year-old patient with no known primary metabolic disease (BMI 20.8 kg/m2). Compartments of subcutaneous fat tissue (not labeled here, area 89.3 cm2) and visceral fat tissue (yellow, 43.5 cm2), psoas muscle (green, 12.6 cm2) and paravertebral muscles (blue, 50.1 cm2). Mean muscle attenuation is 40.4 Hounsfield units (HU).
CT plays a special role in the context of gathering image data for BC analysis. CT imaging is relatively fast and uncomplicated, often part of a routine radiological examination, and also performed repeatedly during follow-up. [Fig. 3] shows a sample case for the evaluation of a patient with a combination of sarcopenia and obesity. One advantage of CT over MRI is the standardized scale for measurements (Hounsfield units). In the future, quantitative CT could also play a greater role in the characterization of MASLD [52]. There is, for example, a deep learning analysis of the skeletal muscles, which is based on routine CT scans of the abdomen [53]. Moreover, DL algorithms can also be used for quality control of the imaging itself [54].
Fig. 3 Mean muscle (radiation) attenuation (MA) for evaluating sarcopenic obesity. Axial CT images of two female patients with a similar BMI (A: 32.3 kg/m2, B: 33.5 kg/m2) but different MA (A: 47.2 HU, B: 11.3 HU).
In recent years, MRI-based BC analyses have improved the characterization of important metabolic diseases, including the metabolic syndrome and type 2 diabetes. The focus has been on the association with insulin resistance, visceral fat tissue, and treatment success after obesity surgery. [Fig. 5] shows an example of a longitudinal MRI examination of relative fat content before and after obesity surgery. MRI has also been increasingly used for pharmacological studies. PDFF measurements, for example, have demonstrated, at least preliminarily, a positive effect of semaglutide on MASLD; in 2023, the drug has also been publicly referred to as “weight loss medication” [55]
[56]. Averaged PDFF values in the liver were also found to decrease after bariatric (obesity) surgery ([Fig. 6]).
Fig. 5 Abdominal fat distribution of a 54-year-old man at the level of the navel before as well as 6 and 12 months after bariatric surgery (Roux-Y bypass). The visceral fat volume determined after segmentation of all axial MR slices (between diaphragm and symphysis) was significantly reduced – 8.3 L > 4.9 L > 4.2 L.
Fig. 6 Proton density fat fraction (PDFF) maps of the liver prior to obesity surgery, Roux-Y gastric bypass (average over shown region of interest 14 %) and in the follow-up examinations after 1 and 7 months (8.5 % and 4.6 %, respectively).
The MRI sequences needed for fat quantification (T1-weighted or Dixon technique) are available on nearly every system. So far, however, the method is not part of the routine examination protocol [57]. Interactive segmentation of fat and muscle tissue takes time and training, but many manufacturers are already providing advanced application modules to automatically evaluate and visualize the results. There are actually reports about corresponding requests to clinical radiologists. In addition, there are some commercial providers that offer such a non-clinical service.
The training of deep learning models is substantially more difficult in the case of MRI images due to the multitude and variability of image weighting with sequences being selected as a function of anatomical region and clinical question. Conventional T1-weighted images may also be used for quantification even though segmentation is slightly more demanding. The advantages of using thicker slices, i. e., less efforts for acquisition and segmentation of the images and a higher signal-to-noise ratio, usually outweigh the disadvantage of a lower spatial resolution. For practical reasons and due to time constraints, many MRI studies, like their CT counterparts, restrict their analysis to single (representative) slices. An individual comparison with the total abdominal volumes showed, for example, a relatively good linear correlation for various fat compartments [19]
[20]
[21]
[22]. The MRI sequences on most systems have become so time-efficient that even whole-body imaging takes only a couple of minutes. [Fig. 4] illustrates the use of MRI for the evaluation of patients with obesity.
Fig. 4 MRI examination of patients with obesity: Comparison of two patients with a similar age and BMI (A: female, 50 years old, BMI 40.9 kg/m2, B: male, 48 years old, BMI 39.2 kg/m2). Initial cross-sectional imaging (A: at the level of the middle lumbar spine, L3 / L4, B: at the level of the lumbosacral joint L5 / S1) shows the differences between gluteofemoral (A) and visceral (B) fat distribution.
In PET-MRI whole-body imaging, Dixon sequences are routinely used for attenuation correction so that corresponding fat and water images are already available. Such datasets can now be segmented fully automatically to analyze compartments like fat, muscle, bone, and organs, also in pediatric patients [58]. New PET detectors allow multiphase PET/CT examinations during continuous table movement with robust data transfer (better than the conventional step-and-shoot technique) and are used, for example, in oncological imaging. Such technological progress should gradually provide deeper insights into specific metabolic processes. Time-resolved methods, e. g., dynamic whole-body PET/CT, are currently not widely available [59].
A tomographic assessment of the body composition is closely related to the clinical question, which is why serial examinations are rare. In one example, a mobile MRI system was used to determine body composition over the course of an ultramarathon (4500 km). The athletes showed a significant decrease in muscle and fat mass (VAT and SAT) [60].
The results of user-dependent BC analyses should be interpreted with some caution [61]. With respect to accuracy and reproducibility, some deep-learning approaches now outperform conventional segmentation of selected slices [62]. However, there are no standards with which the available evidence can be transferred to the clinical routine. This includes clear requirements, particularly in the case of therapeutic questions, e. g., for defining a controlled physical activity during sport or activity interventions. Another critical issue are the specific hardware components used. For example, what is the potential impact of a particular imaging system or imaging protocol? One of the few methodological studies on 18 subjects showed, for example, that the inaccuracy of BC analyses by MRI (including PDFF in the liver, fat, and muscle volume) was largely determined by the repeatability of the measurement on the same system [63].
Current developments and outlook
Current developments and outlook
In recent years, radiological methods have been increasingly used for BC analyses, particularly in sarcopenia [64]
[65]. Disregarding a few exceptions, low muscle mass is now widely considered a risk factor for an unfavorable course of chemotherapy. The basis for this finding is a meta-analysis from the year 2023, which included 35 studies and over 3800 patients [66]. In 2022, the results of an automatic BC analysis were published with over 9200 asymptomatic adults undergoing colon cancer screening by CT over a median time period of 9 years. The X-ray attenuation of the skeletal muscles and the calcium content of the abdominal aorta had a prognostically significant effect on the 10-year survival rate. The AUC (area under the curve) was 0.72 (men) and 0.76 (women) [67].
Some fundamental limitations remain. For example, a standard for collecting and processing BC data is often lacking. The analyses are performed anyway, sometimes simply because many parameters are relatively easy to measure. Radiological BC parameters are often measured in clinical departments whose patients undergo radiological imaging. Subsequent correlations with clinical outcome variables are common and range from degenerative orthopedic diseases like lumbar disc herniation [68] to various malignancies (e. g., renal cell carcinoma [69] or non-metastasized colorectal carcinoma [70]), inflammatory diseases (chronic inflammatory bowel diseases) [71], and hospitalization in SARS-CoV2 infection [72]. BC analyses are usually aimed at a specific question and population and in-depth assessments are often secondary [4]. BC parameters can also be derived from the non-diagnostic CT information of a planning CT scan [73]. At the other end, controlled analyses from large cohorts like the UK Biobank
[74] or National Cohort (Germany) [49] are needed for a reliable translation into the clinical routine but remain rare.
Considering the lack of standards and reference values, there is a demand for more evidence for radiological BC analyses. Already in 2007, the Quantitative Imaging Biomarkers Alliance (QIBA) of the Radiological Society of North America (RSNA) was founded with the goal of achieving the highest possible validity and reproducibility in radiological image data analysis. One prominent result is the multicenter evaluation of multi-echo MRI sequences (Dixon technique), the basis for determining liver fat content (PDFF) [75]. It was shown that field strength, MRI manufacturer, and reconstruction method barely had an influence on reproducibility. The clear methodological specifications were a pivotal element of this study. The increasing demand for scientific publication in combination with easily available image data may potentially provide some room for less stringent evaluations. In radiology, there is a growing demand to ensure that the evidence of such analyses has a solid basis.
Essentially, radiological imaging is meant to answer specific questions from the referring physicians (after justification of the indication) to ultimately provide optimal care to the patient. Reviewing the indication for a CT examination is particularly strict because patients will be exposed to ionizing radiation. Modern multidetector CT scanners generate high-resolution 3D data sets with complete body coverage, e. g., in routine staging examinations. Regardless of the original indication, the image data can be used for “opportunistic screening”, i. e., a search for further diagnostic findings not yet reported [76]. It is then possible to derive quantitative metabolic or BC information from these data sets that may ideally serve as a biomarker. Important imaging features include the mineral salt content in osteopenia, visceral fat volume in overweight and obesity, vascular calcifications in arteriosclerosis, intrahepatic fat in MASLD, and skeletal muscle mass and quality in sarcopenia. From a radiological standpoint, there is a clear medical advantage for patients. Outside the context of a clinical or scientific study, however, the reimbursement of these additional services needs to be clarified [77].
An individual risk profile can be defined from the wealth of features. BC measurements are also of interest from an ethical point of view. If, for example, the screening examination does not show the wanted imaging feature, a quantitative BC analysis may provide additional information. The question that arises here is whether potentially relevant BC information that is rather easy to assess can actually be withheld from the patient? The extra efforts would already be acceptable today but in the future, such evaluations will likely become automated. Some experts already consider BC information as a contribution to a radiology, where the value of information is more important than its volume [53].
The number of examinations and data volume in radiology will likely increase. Traditional, risk-stratified, step-by-step diagnosis could be replaced by a higher prioritization of imaging, particularly when disadvantages like radiation exposure are further reduced by technology, and referring physicians, e. g., in the emergency department, demand a higher efficiency. A relative lack of medical expertise and insufficient training structures could reinforce the trend to verify clinical decisions with a broader indication for imaging. So far, urgent, i. e., early imaging has been reserved for emergencies, e. g., of the cardiovascular system. The indication for imaging over other diagnostic tests is already discussed in the literature: what is the effect of a CT examination on the clinical course in older people with a suspicion of acute appendicitis [78]
[79]
[80]? If cross-sectional imaging would be performed at the beginning of the diagnostic chain, a BC analysis with metabolic risk profile could be useful for more individualized patient care. This would require well-trained personnel, modern and available imaging systems with a corresponding IT infrastructure, and regulated allocation of costs.
It can be concluded that the value of radiology for the evaluation of body composition has increased in recent years. The application spectrum will probably grow as a result of the fundamental challenges faced by the health care system, particularly the demographic change and the increase in obesity-related diseases. For the field of radiology, it is therefore important to keep an eye on the overall picture and the relevant trends and interactions between the players involved.
