Rethinking SUV Normalization in PET: A Physiology-Guided Approach over the current Geometrical Approach
In PET imaging, the Standardized Uptake Value, or SUV, has long been the metric used for interpreting tracer uptake. Its appeal lies in its simplicity: it offers a normalized measure of how much tracer is retained in a region of interest, scaled to the injected dose and patient size, usually body weight, surface area, or lean mass. Yet beneath this simplicity are assumptions that can distort what SUVs reflect on tissue metabolism, disease activity, or therapeutic response.
One fundamental flaw in traditional SUV construction is the assumption that tracer distributes uniformly across the body (injected dose per body weight), or across the muscle compartments (injected dose/lean body mass). But this is rarely accurate, as tracer uptake is governed primarily by patient physiology, not geometry. It’s driven by regional blood flow, membrane transport, receptor density, and tissue-specific metabolic processes, all of which can vary dramatically across individuals and disease states. Even in individuals with identical body weight or lean body mass, the actual distribution of a tracer like FDG or DOTATATE may differ greatly due to age, inflammation, tumor burden, organ function, or treatment effects.
Additionally, conventional SUV normalization methods use population-derived formulas to “estimate” lean body mass ignoring that actual muscle mass and fat distribution can vary significantly, especially in patients with cancer, cachexia, obesity, or sarcopenia.
What’s missing here is a data-driven understanding of how the injected tracer truly distributes within each individual. That’s where modern automated whole-body segmentation platforms, such as DAFS, capable of highly accurate and reliable automated segmentation of the entire body CT scan, offer a new opportunity for quantitative data-driven normalization of PET measurements.
Rather than normalizing tracer values based on geometrical assumptions, we can instead calculate a patient-specific model of where the tracer is distributing, as a function of space, and time, driven by variations in tracer movement across the body. By segmenting all organs and tissues where the tracer is physiologically expected to accumulate, such as the brain, heart, lungs, liver, spleen, skeletal muscle, bone, and tumors, and excluding regions that are primarily excretory or passive conduits, like the bladder, bowel, and large blood vessels, we can define a relative tracer distribution volume: the fraction of the body that is actively participating in tracer metabolism.
This individualized distribution model can then be used to normalize tracer uptake based on actual physiology, not just body geometry. It accounts for how much of the body’s tissue mass is truly engaged in the tracer’s pharmacokinetics, a value that can differ between patients of the same weight, or even in the same patient over time, before and after treatment. By incorporating dynamic PET data, this normalization model can also evolve with time during a scan, capturing how tracer distribution patterns change at early versus late imaging frames of acquisition where in the early frames, uptake is dominated by perfusion-heavy organs (heart, lung) whereas in the late frames, uptake reflects metabolic trapping (muscle, tumor, marrow).
Furthermore, acknowledging, for example, that the liver consistently retains more FDG than skeletal muscle, or that bone marrow may become more active after chemotherapy, this approach can incorporate weighting based on average metabolic activity. These physiological weighting factors, derived from reference cohorts or population data, allow the normalization framework to reflect not just the shape of the body, but also its metabolic architecture.
The impact of this refined normalization could be potentially significant. It allows more reliable comparison of PET scans across individuals, particularly in clinical trials or multi-center studies. It also enhances longitudinal consistency, enabling more accurate assessment of response to therapy over time. Most importantly, it grounds PET quantification in actual biology, improving diagnostic accuracy and reducing misleading variability introduced by geometrical body metrics.
By moving SUV normalization from an era of approximation based on geometry and body weight/size into one incorporating physiology extracted from data, this may open the door to more meaningful interpretations of PET data, and, ultimately, to a more accurate, quantitative and personalized applications of nuclear imaging in clinical workflows.