At the heart of the argument is a deceptively simple observation: a single computed tomography scan performed for an entirely unrelated reason—an abdominal emergency, oncologic staging, or presurgical planning—already contains thousands of volumetric images dense with quantitative information about the patient’s body. CT works by passing an X-ray beam through body tissue and collecting the attenuated signal with detectors, producing images in which each pixel carries a densitometric value expressed in Hounsfield units. Because each tissue type attenuates X-rays in a characteristic way, researchers can apply density thresholds to separate bone, skeletal muscle, subcutaneous adipose tissue, and visceral adipose tissue, then measure them precisely. A single axial landmark at the level of the third lumbar vertebra has emerged as the standard reference point, since at this level the abdominal musculature, psoas and paravertebral muscles, and both fat compartments can all be assessed. This concept of “opportunistic screening”—extracting body composition data retrospectively from exams ordered for other clinical questions—could, the authors argue, significantly aid large-scale health prevention without exposing patients to any additional radiation or cost.

The clinical stakes are considerable. The review details how body composition parameters function as biomarkers across a five-level model spanning atomic, molecular, cellular, tissue, and whole-body scales, with the three-compartment model of fat, lean mass, and bone offering the most informative practical framework. Fat, far from being a homogeneous reservoir, comprises subcutaneous and visceral depots with sharply divergent metabolic consequences. When subcutaneous storage capacity saturates, excess fat accumulates viscerally and then ectopically in the liver, heart, muscles, and skeleton. Increased pericardial fat has been linked in a meta-analysis of 83 studies to coronary artery disease and atrial fibrillation, while intrahepatic fat drives insulin resistance, systemic inflammation, fibrotic change, and elevated hepatocellular carcinoma risk within metabolic dysfunction-associated steatotic liver disease. Perirenal fat accumulation affects glomerular filtration rate and chronic kidney disease risk through both mechanical effects on neighboring vascular and lymphatic structures and systemic cytokine production. Even the familiar body mass index, the review notes, cannot distinguish lean mass from fat mass, underscoring the need for more refined, clinically practical tools.

Skeletal muscle occupies an equally central position in the analysis. Accounting for roughly forty percent of total body weight, muscle mass is influenced by nutritional status, physical activity, endocrine milieu, and disease, and it undergoes complex age-related changes involving type-II fiber atrophy, low-grade inflammation, and motor-unit loss through denervation and reinnervation. These processes produce clinically distinct entities: myopenia, characterized by low muscle mass, and sarcopenia, defined as low muscle strength confirmed by reduced mass or muscle attenuation and graded by poor physical performance. Sarcopenia, whether primary and age-related or secondary to chronic diseases such as cancer and heart failure, is associated with increased risks of falls and fractures, disability, and mortality, particularly in older adults and oncologic patients. The review also highlights cachexia, a complex metabolic syndrome driven by underlying illness and systemic inflammation that cannot be fully reversed by conventional nutritional support, and which correlates with reduced treatment tolerance and survival. Muscle evaluation, in short, is not an academic exercise but a direct determinant of treatment decisions and outcomes.

Bone completes the triad. The review distinguishes carefully between bone mineral content, expressed in grams as the skeleton’s mineral fraction, and bone mineral density, expressed as a concentration in grams per square centimeter, which estimates skeletal strength. The two measures are related but not interchangeable, and declines in the bone compartment signal loss of osseous mass, altered microarchitecture, and remodeling imbalance linked to elevated fracture risk, poor quality of life, and increased mortality. Crucially, the three compartments do not operate in isolation. Skeletal muscle supports bone through mechanical loading and biochemical signaling, while adipose tissue influences bone through adipokines and marrow fat content, all interconnected via endocrine, inflammatory, and biomechanical pathways. Pathological alterations can therefore coexist as osteosarcopenia, or with excess adiposity as osteosarcopenic obesity, producing greater frailty, falls, and fracture risk than any single deficit alone—evidence, the authors contend, for concurrent assessment and integrated intervention.

Against this clinical backdrop, the review surveys the imaging modalities that make quantitative assessment possible. Dual-energy X-ray absorptiometry remains the most widely used technique, differentiating fat, lean soft tissue, and bone by measuring tissue attenuation of X-rays emitted at two energy levels. But DXA carries a critical assumption: that lean soft tissue is approximately seventy-three percent water. When hydration deviates—in edema, ascites, inflammatory states, heart, renal or liver failure, or after recent endurance exercise—the extra water inflates the lean mass estimate, reducing accuracy and reproducibility. Fluid-overloaded patients may require repeat measurement in a euvolemic state, with CT or MRI-based assessment as a complement less sensitive to hydration bias. MRI offers powerful quantitative fat-water imaging techniques such as Dixon sequences and proton density fat-fraction reconstruction, but lacks the fixed signal-intensity thresholds that make CT segmentation straightforward, and suffers from high cost and incompatibility with certain metal implants.

This is precisely where artificial intelligence enters the picture, and where the review’s core contribution lies. Manual segmentation of a single CT slice is time-consuming, poorly repeatable across readers, and unstandardized, and volumetric datasets multiply the burden impossibly. AI methods trained on large annotated datasets offer automated, highly reproducible segmentation that eliminates the bottleneck of manual or semi-automatic delineation—automatic muscle segmentation for sarcopenia assessment being the flagship example. Beyond segmentation, the review describes AI models pursuing more complex goals: integrating imaging features with clinical and molecular biomarkers to support translational, personalized risk assessment and treatment planning. In this vision, an AI system evaluating a CT scan ordered for an unrelated reason could opportunistically flag sarcopenia, quantify visceral fat, estimate bone density, and stratify the patient’s cardiovascular and metabolic risk, all while the radiologist attends to the original clinical question.

The oncological implications receive particular emphasis. Sarcopenia and adipose tissue distribution are recognized as key risk factors for overall survival, postoperative complications, and treatment-related toxicity in cancer patients. The review cites evidence that sarcopenia substantially increases perioperative mortality, with one synthesis of 42 studies reporting an odds ratio of 2.40, and notes that socioeconomic disadvantage is itself associated with body composition measurements crossing thresholds predictive of cardiovascular and metabolic mortality. Automatic body composition analysis during follow-up imaging enables quantitative monitoring of tissue changes over the disease course, potentially allowing clinicians to adjust nutritional, exercise, and pharmacological interventions before irreversible functional decline occurs. The authors frame this as the practical realization of personalized medicine: treatment decisions informed not by a single number on a scale, but by a dynamic, quantitative portrait of the patient’s internal anatomy.

The review is candid about limitations. CT involves ionizing radiation and high cost, precluding its use purely for screening; MRI remains expensive and inaccessible to patients with non-compatible devices; AI models depend on the quality and diversity of their training data and must demonstrate robust performance across scanners, protocols, and populations before routine deployment. Standardization of measurements, validation of predictive models in prospective trials, and integration into clinical workflows all remain unresolved challenges. The authors also caution that AI-generated measurements must be interpretable and actionable for clinicians who may have no radiological training, which is why the review addresses its message not only to radiologists but to the broader clinical community.

Nevertheless, the trajectory described is unmistakable. The authors conclude that the synergy between artificial intelligence and body composition analysis can enhance the management of health conditions from diagnosis to personalized treatment, and that opportunistic extraction of body composition data from existing imaging represents one of the most efficient prevention strategies available to modern health systems. As AI tools mature and validation studies accumulate, the modest lumbar spine cross-section captured on an ordinary abdominal CT may become one of medicine’s most information-rich real estate—a few square centimeters of pixels from which algorithms can read a patient’s metabolic future. What was once the domain of specialized research laboratories, the review suggests, is rapidly becoming an expectation of standard care.