When analyzing medical images, raw numbers can be deceiving. A pixel's brightness in a standard MRI scan or the raw photon count in a nuclear medicine scan does not correspond to a fixed physical quantity. Instead, these values fluctuate based on scanner hardware, patient positioning, and software configurations.
To make sense of these images, scientists and clinicians traditionally rely on relative normalization. This technique scales the intensity of an image by comparing target areas to an internal baseline—either a designated "healthy" reference region or the global average of the entire organ.
While this approach works well for localized injuries like a stroke or a focal tumor, it introduces a severe diagnostic blind spot when dealing with systemic or widespread conditions. When a disease alters the entire organ uniformly, the internal baseline shifts along with the pathology. As a result, the relative differences disappear, and the imaging software can mistakenly classify a deeply diseased organ as perfectly healthy.
The Illusion of the Uniformly Healthy Organ
To understand why relative normalization fails, consider a cardiac phenomenon known as balanced ischemia. In coronary artery disease, clinicians frequently look for localized drops in blood flow to identify blocked vessels. If one artery is narrowed, the tissue it supplies will appear darker than the surrounding healthy tissue.
However, if a patient suffers from severe, balanced multivessel disease, all major coronary arteries are narrowed to a similar degree. As Di Carli et al. explain, when blood flow drops uniformly across the entire heart, relative tracer distribution methods fail to show any comparative deficits. Because every region of the heart muscle is equally impaired, the internal comparisons used by standard imaging software make the heart appear uniformly healthy.
Localized Disease (Visible) Balanced Global Disease (Hidden)
[ Healthy ] [ Healthy ] [ Impaired ] [ Impaired ]
[ Healthy ] [ Blocked ] [ Impaired ] [ Impaired ]
^ relative contrast ^ NO relative contrast
reveals pathology (appears uniformly normal)
As Ben-Haim et al. note, this creates a dangerous diagnostic blind spot. Because standard cardiac stress tests rely on comparing tracer uptake between different regions, they lack a true baseline when circulation is reduced globally. Arash Kardan demonstrated that standard visual interpretations of these scans regularly miss critical, high-risk conditions because they can only identify relative differences rather than absolute deficits.
The Moving Anchor in Neuroimaging
This normalization bias is not unique to cardiology; it presents an identical mathematical challenge in neurology. When imaging neurodegenerative or neuroinflammatory diseases, researchers often normalize brain scans using a "pseudo-reference" region, such as the cerebellum, which is assumed to be unaffected by the disease.
If this assumption is wrong, the entire analysis degrades. Kagitani-Shimono et al. showed that when a disease actually involves the chosen reference region, the resulting signal-to-noise ratio is severely underestimated, masking significant pathological changes. Similarly, Wiltgen et al. emphasize that because systemic neurological diseases can alter the entire brain, relying on internal tissue as a reference introduces systematic scaling biases that obscure the very biomarkers researchers are trying to measure.
The problem deepens when using standard statistical normalization techniques like Z-score standardization, which scales data based on the global mean and variance of the image. As McGinnis et al. point out, when a disease drives a global change across the entire brain volume, those changes alter the reference mean and variance itself. Mathematically, the Z-score calculation ends up absorbing and erasing the biological signal, effectively wiping away the relevant clinical differences between patient groups during data preprocessing.
Why the Mathematics Erase the Signal
The core mechanism of this failure can be captured by looking at how normalization software processes an image. Early quantitative tools, such as the polar maps evaluated by Sasaki et al., were designed to automatically scale an organ's imaging values against its highest-performing region.
Mathematically, if $I(x)$ represents the raw intensity of a pixel and $I_{\text{max}}$ represents the patient's healthiest reference region, the normalized pixel value $N(x)$ is calculated as:
$$N(x) = \frac{I(x)}{I_{\text{max}}}$$
In a healthy patient, $I_{\text{max}}$ is high, and normal variations across the organ create a predictable baseline map.
In a patient with global pathology, every single pixel value drops. If the true physiological function drops by 50% across the entire organ, then both the target pixel $I(x)$ and the reference pixel $I_{\text{max}}$ are cut in half:
$$N_{\text{diseased}}(x) = \frac{0.5 \cdot I(x)}{0.5 \cdot I_{\text{max}}} = \frac{I(x)}{I_{\text{max}}} = N_{\text{healthy}}(x)$$
Because the disease scales both the numerator and the denominator equally, the normalized output $N(x)$ remains completely unchanged. The software incorrectly identifies the compromised state as normal because the relative difference falls below detection thresholds.
The Solution: Absolute Physiological Quantification
To break free from the limitations of relative normalization, the field of medical imaging is shifting toward absolute physiological quantification. Instead of asking how one region compares to another, advanced imaging modalities use dynamic data acquisition and kinetic modeling to measure real physical properties in standardized units, such as milliliters of blood flow per gram of tissue per minute ($\text{mL/min/g}$).
Relative Assessment Absolute Quantification
[ Region A ] vs [ Region B ] [ Region A ] = 0.6 mL/min/g
(Tells you only if A is (Tells you exactly how much
better or worse than B) blood is flowing, regardless of B)
By measuring precise flow per unit of tissue, absolute quantification completely removes the need for an internal healthy reference region. As Sohn et al. explain, this representational strategy allows clinicians to detect global heart disease and balanced ischemia that relative perfusion imaging completely misses.
A consensus statement by Dewey et al. underscores that transitioning from relative comparisons to absolute blood flow measurements provides a fundamentally more accurate physiological assessment. When clinicians can measure the true, unscaled physiological output of an organ, they can confidently diagnose widespread, uniform pathologies that have historically hidden behind the illusion of relative balance.
Key Foundational Work
62. Assessment of Detection Ratio of Ischemic Heart Disease using Diagrams of Polar map (1990) — contributed 3 concept elements to this direction.
Clinical Myocardial Perfusion PET/CT (2007) — contributed 2 concept elements to this direction.