Hounsfield units (HU) are the density numbers on a CT. Water is 0 HU. Air is −1000 HU. Everything else is attenuation relative to water. This page is the scale. It is not a windowing how-to.
If you meant the two sliders (WW vs WL) → window width vs window level. If you meant apply WC/WW in Python → CT scan window settings.
The two anchors
Every scanner is calibrated to the same two points, so a liver in one room is supposed to read like a liver in another.
- Distilled water = 0 HU. Denser than water → positive, brighter. Less dense → negative, darker.
- Air = −1000 HU. The low end of the body-scale.
The scanner measures how much of the X-ray beam a voxel absorbs (attenuation) and maps that to HU with those two goalposts. Bone is a wooden door; fat and lung are closer to glass. Sir Godfrey Hounsfield’s 1971 scanner is why the unit has his name. History of the machine itself is not this page.
Where tissues land
| Substance or tissue | Typical HU |
|---|---|
| Air | −1000 |
| Lung | −700 to −500 |
| Fat | −100 to −50 |
| Water | 0 |
| CSF | +8 to +15 |
| Kidney | +20 to +40 |
| Soft tissue | +30 to +80 |
| Muscle | +35 to +55 |
| Liver | +50 to +70 |
| Acute blood | +55 to +75 |
| Calcification / stone | +100 to +400 |
| Cortical bone | +1000 or more |
A region-of-interest tool returns these numbers, not a vibe. Fat in the abdomen near −90 HU is fat. A bright speck in a kidney at +400 HU is a stone against parenchyma around +30 HU.
Cyst, tumor, bleed
- Cyst vs solid mass. A simple cyst tracks water: about 0 to +20 HU. A solid tumor is cellular and denser: about +40 to +80 HU. That one measurement decides “watch” vs “work up.” It does not diagnose cancer by itself.
- Acute bleed. Fresh clot is about +55 to +75 HU — brighter than liver or spleen next to it, which is why a trauma CT can call hemorrhage without waiting for a later phase.
HU is a tip, not a verdict. History, the rest of the study, and often a biopsy still sit on top of the number.
Why the number moves
The scale is stable. The reading is not a steel ruler.
- Beam hardening. Dense bone or metal eats the soft part of the beam. What remains is “harder.” Nearby soft tissue can read darker than it is. That is a caveat on the number, not the artifact catalog — streaks, cupping, and MAR live on artifacts in computed tomography.
- IV contrast. Iodine is supposed to raise HU in vessels and enhancing tissue. Compare pre and post; do not treat a post-contrast liver as a native +50 to +70.
- Habitus and calibration. Body size and a scanner’s daily air-cal nudge the reading a little. Radiologists read inside that tolerance, not against a fake exact.
Windowing, one paragraph
A CT stores thousands of HU. A screen shows about 256 greys. Windowing picks a band of that scale and stretches it: a bone window parks on high HU, a lung window parks on very low HU. The sliders are window width and window level — that comparison is window width vs window level. Applying a pair to a series in Python is CT scan window settings. This page stops at the scale.
Do all scanners agree? Close enough. Water stays 0 and air stays −1000 by calibration. Age, kV, and the morning air-cal introduce small drift. That is expected; it is not a reason to invent a second scale.
PYCAD builds products that have to treat HU as a number, not a shade of grey. Case studies.