The Hounsfield Unit Chart: What the Numbers Mean

Every pixel on a CT scan is assigned a number that reflects how much the tissue in that spot slowed down the X-ray beam, and that number is expressed in Hounsfield units. The scale is anchored by two fixed points: water is defined as 0 HU and air as −1,000 HU. Everything else falls somewhere on that continuum based on its density and atomic composition, from airy lung tissue deep in the negatives to dense cortical bone well above +1,000. Understanding what those numbers mean turns a grayscale image into something a clinician can actually measure, compare, and use to make decisions.

How the Scale Is Built

The Hounsfield unit scale is named after Godfrey Hounsfield, the British engineer who developed the first CT scanner in the late 1960s. The idea was straightforward: fire X-ray beams through the body at many angles, measure how much each beam was weakened, and use a mathematical algorithm to reconstruct a two-dimensional image of the interior.

Water sits at zero because it serves as the universal reference material. Anything denser than water pushes into positive territory; anything less dense drops into the negatives. Dense cortical bone typically measures around +1,000 HU, while the air in your lungs sits near −1,000 HU. The original scale ran from −1,024 to +3,071, a range of 4,096 values encoded in 12 bits. Modern scanners can extend that range significantly, which matters when metal implants are involved.

Where Common Tissues Fall on the Chart

Fat is one of the easiest tissues to recognize on CT because it consistently reads well below zero. A study deriving reference intervals for abdominal fat found visceral fat falling roughly between −122 and −84 HU and subcutaneous fat between −123 and −93 HU.¹ A separate population study reported that average visceral fat attenuation ranged from about −90 to −101 HU depending on the spinal level measured and whether contrast was used.² The consistency of fat’s negative reading is what lets software automatically segment it from surrounding tissue.

Lung tissue at full inspiration normally measures between roughly −750 and −850 HU. That range reflects the enormous amount of air mixed with thin-walled tissue in healthy alveoli. When lung parenchyma drops below about −900 HU, it crosses a threshold long used to identify emphysema, a condition where alveolar walls have been destroyed and air trapping has increased.³ In other words, lungs that read “too airy” on the Hounsfield scale are flagged as potentially damaged.

Most non-fatty soft tissues, such as muscle, liver, spleen, and kidneys, cluster between roughly +20 and +70 HU on an unenhanced scan. This narrow range is both useful and limiting. It means clinicians can spot abnormalities when a tissue’s density drifts outside its expected band, but it also means small differences between organs can be hard to see without contrast enhancement.

Fresh, unclotted blood generally reads around +30 to +45 HU. Acute hemorrhage, where blood has clotted and concentrated, reads higher, often +50 to +70 HU or more. As a hematoma ages, its density changes: a study tracking intracerebral hemorrhage over time found that mean hematoma density increased between baseline and follow-up scans, while signs of internal heterogeneity decreased.⁴ This evolving density is one reason radiologists can roughly estimate the age of a bleed.

Cortical bone usually falls between +700 and +3,000 HU depending on location and the patient’s bone health. Cancellous (spongy) bone inside vertebral bodies reads much lower, and that lower reading becomes clinically meaningful for detecting osteoporosis, as discussed below.

Kidney Stones and What Their Density Reveals

One of the most practical uses of HU values outside the body’s soft tissues is characterizing kidney stones. Because different stone compositions absorb X-rays differently, their Hounsfield readings offer a noninvasive clue to what a stone is made of. CT-measured density has been used to predict stone type and guide treatment decisions.⁵

Calcium-based stones, the most common type, tend to read highest. Research comparing stone types found that no non-calcium stone exceeded about 448 HU, meaning that a stone measuring above that range is very likely calcium-containing.⁶ A study using a core-density measure also found it could distinguish calcium stones from struvite stones with good accuracy at a cut-off of 35 HU per millimeter.⁷ Uric acid stones, by contrast, tend to read much lower, often below 500 HU. This matters because uric acid stones can sometimes be dissolved with medication, while calcium stones typically cannot. A high HU reading, in other words, can steer a urologist toward surgical removal rather than a trial of oral therapy.

The 10-HU Rule for Adrenal Masses

When a CT scan incidentally finds a lump on an adrenal gland, the first question is usually whether it is a benign adenoma or something more concerning. For decades, the working rule has been that a mass measuring 10 HU or below on an unenhanced scan is almost certainly a lipid-rich adenoma, because that low density signals intracellular fat.⁸

The trouble is that a strict 10-HU cutoff misses a lot of adenomas. One study of adrenalectomy patients found that the 10-HU threshold caught fewer than half of confirmed adenomas, with a sensitivity of only about 48%, even though its specificity was above 93%.⁹ Raising the cutoff to 16 HU roughly doubled sensitivity to about 65% without sacrificing specificity. A separate study in adrenalectomy patients similarly found that bumping the threshold to 20 HU yielded a sensitivity of 92% and specificity of 82%.¹⁰ The clinical takeaway is that the classic 10-HU rule is conservative. It is very good at confirming an adenoma when met, but many genuine adenomas have lipid-poor interiors that push their readings above 10.

Screening for Osteoporosis Through Existing CT Scans

Bone mineral density is traditionally measured with a dedicated DXA scan, but many patients have abdominal or spinal CT scans done for other reasons, and the HU values of their vertebral bodies carry useful information. Healthy vertebral cancellous bone reads in the range of roughly 150 to 250 HU. As bone mineral is lost, those numbers drop. Researchers have worked to establish HU thresholds that flag likely osteoporosis so that patients who already had a CT do not need a separate DXA study.

One study of patients with lumbar degenerative disease found that defining osteoporosis as L1 at or below 110 HU, L2 at or below 100 HU, L3 at or below 85 HU, or L4 at or below 80 HU captured about 61% of DXA-confirmed osteoporosis cases at roughly 89% specificity.¹¹ Another group calculated that an average L1–L2 HU cutoff of 110 was about 91% sensitive and 89% specific for osteoporosis when validated against quantitative CT bone density measurements.¹² In patients with ankylosing spondylitis whose spines have fused into a “bamboo” configuration, DXA often overestimates bone density because the fused bone surface misleads the measurement. Using an HU cutoff below 135 identified far more osteoporosis cases than DXA alone in these patients.¹³

This kind of “opportunistic screening” is appealing because it adds zero radiation and zero cost to a scan that was already ordered. A radiologist simply measures the vertebral body and flags patients who fall below the threshold for follow-up.

Why the Same Tissue Can Read Differently on Different Scanners

A persistent problem with Hounsfield units is that they are not perfectly standardized across manufacturers or even across scan protocols on the same machine. A study that scanned the same patients on two different multidetector CT scanners found statistically significant differences in HU measurements at every measured anatomic site. The consistency between the two machines was low in some locations, with subcutaneous fat in one area showing particularly poor agreement.¹⁴

Part of this variability comes from the X-ray energy used. Changing the tube voltage (the kVp setting) has little effect on soft tissue HU values, but high-density materials like titanium and stainless steel show a marked drop in their CT numbers at higher kVp settings.¹⁵ Beam hardening adds another layer: when the X-ray beam passes through a thick or dense body part, lower-energy photons are preferentially absorbed, shifting the effective energy of the remaining beam. One study found that while soft-tissue HU values were stable despite added attenuation, bone readings dropped noticeably, and the magnitude of the drop correlated with the overall tissue thickness the beam had to traverse.¹⁶

For most diagnostic imaging this variability is manageable. Radiologists know the ballpark for each tissue and adjust their judgment accordingly. But it becomes a serious engineering challenge in radiation therapy planning, where an HU value is converted into an electron-density figure that directly determines how much radiation dose a tissue absorbs.

How Contrast Dye Shifts the Numbers

Intravenous contrast agents contain iodine, which absorbs X-rays heavily and drives HU values upward in any tissue the contrast reaches. A blood vessel that reads +40 HU on an unenhanced scan may jump well above +200 HU during peak contrast. The liver typically brightens by +40 to +60 HU or more, depending on the phase of the scan. This enhancement is what makes contrast CT so useful for spotting tumors (which often have abnormal blood supply) or vascular problems.

One question clinicians have asked is whether the concentration of iodine in the contrast vial matters if the total iodine dose delivered is the same. A randomized study comparing two iodine concentrations at identical total iodine loads and injection times found no significant difference in the degree of enhancement of the aorta, portal vein, or liver at most scan phases.¹⁷ What drives enhancement, in short, is total iodine delivered, not how concentrated the bottle is.

This also means that any HU-based diagnostic threshold designed for unenhanced scans, like the adrenal adenoma cutoffs discussed earlier, becomes unreliable after contrast is given. An adenoma that would measure 8 HU on a plain scan may read 50 HU or higher in the arterial phase. Radiologists account for this with washout calculations, measuring how quickly contrast leaves a mass, but the underlying point is that comparing HU values between enhanced and unenhanced scans is not straightforward.

Window and Level Settings Change What You See, Not What Is There

A CT image dataset contains thousands of HU values, but a typical computer monitor can only display about 256 shades of gray. Window and level settings determine which slice of the HU spectrum gets mapped onto those shades. The “level” is the center HU value, and the “width” is how many HU values above and below that center are displayed. Anything above the window appears white; anything below appears black.

This is why the same scan looks drastically different when viewed in “lung window” versus “bone window” versus “soft tissue window.” In a lung window (center around −600, width around 1,500), you can see fine detail in aerated lung tissue, but the mediastinum looks like a uniform white blob. In a bone window (center around +300, width around 1,500), fractures and calcifications pop, but soft-tissue detail vanishes. None of the underlying data changes. The radiologist is simply choosing which part of the Hounsfield scale to spotlight.

A trauma-focused study tested whether a single “all-in-one” window setting could replace toggling between multiple conventional windows. Radiologists using the combined setting were about 14 seconds faster per case with no significant loss in diagnostic accuracy compared to conventional toggling.¹⁸ In a busy trauma bay, those seconds add up, and the result suggests that a broader windowing approach can work when speed matters.

Metal Implants and the Extended Scale

Standard 12-bit CT scales top out at +3,071 HU. Metal hip replacements, spinal hardware, and dental implants often exceed that ceiling, causing the scanner to “clip” their values and produce streak artifacts that corrupt nearby tissue readings. Extended-bit CT reconstruction widens the scale so that metals can be assigned their true high HU values without saturating.

Early work on extended CT scales for radiotherapy patients with metal hip prostheses found that while the extended range could not precisely predict the electron density of every implant material, it could reliably distinguish between different types of implant densities.¹⁹ More recent phantom and patient studies have confirmed that extended-bit reconstruction reduces HU saturation and improves agreement between calculated and measured radiation doses near metal.²⁰ One group concluded that direct dose calculation on extended-scale images in the presence of metal implants can produce accurate treatment plans, eliminating the need for a physicist to manually contour the metal and assign a density override.²¹ As 16-bit reconstruction becomes more widely available, this is increasingly recommended as a default for patients with implants.

Hounsfield Units in Radiation Therapy Planning

In radiation oncology, the CT scan is more than a diagnostic image. It is the physical blueprint for calculating how much dose each tissue will receive. Treatment planning systems convert every HU value into a relative electron density, and that conversion determines how the radiation beam interacts with each voxel of the patient’s body. Getting the conversion wrong means delivering the wrong dose.

The conversion is done through a calibration curve, typically generated by scanning a phantom with inserts of known density. Different phantoms, and different CT scanners, produce slightly different curves. One study comparing calibration curves from three different CT manufacturers found that the differences could affect dose distribution in lung cancer treatment plans.²² The divergence is typically small for soft tissues but can grow at higher densities. A separate study confirmed that HU-to-electron-density calibration curves from commercial phantoms diverge at densities above that of water, and the divergence depends on the elemental composition of the phantom inserts.²³ Automated quality assurance tools are now being developed to regularly verify these curves for different body regions, generating protocol-specific calibrations for head, thorax, abdomen, and pelvis scans.²⁴

Dual-energy CT offers a way to improve HU stability across different scanning conditions. A study comparing twin-beam dual-energy reconstructions with conventional single-energy images found excellent agreement in HU values for major abdominal organs, and the dual-energy images actually showed significantly higher HU stability for most organs, particularly at virtual monoenergetic energies around 80 keV.²⁵ More stable HU values feed into more accurate electron-density conversions, which in turn mean more precise radiation doses.

Post-Mortem Imaging and Density Decay

An unusual but growing application of Hounsfield units is in forensic pathology. After death, tissues decompose and gas accumulates inside the body, and both processes change HU values in predictable directions. A study correlating CT measurements with time since death found that for each additional post-mortem day, HU values of the brain, heart, lung, liver, and spleen decreased, while kidney HU actually increased. Total body gas volume rose steadily over time.²⁶ This kind of data is being explored as a tool for estimating how long someone has been dead without needing an invasive autopsy, which can be useful in disaster victim identification or when religious or cultural considerations make traditional autopsy difficult.

The kidney finding is interesting because it runs counter to the general trend. The authors attributed it to the accumulation of dense decomposition byproducts in renal tissue, though the mechanism is not fully settled. It is a reminder that the Hounsfield scale does not just measure anatomy. It measures whatever physical state the tissue happens to be in at the moment of scanning, living or otherwise.