Hounsfield Units & Windowing
- A Hounsfield Unit (HU) is just a number for "how much did this voxel slow the X-ray beam" — a standardized brightness scale for CT.
- The scale is pinned to two fixed points by definition: water = 0 HU and air = −1000 HU. Everything else is measured against those.
- Rough landmarks: fat is negative, simple fluid is near zero, soft tissue is positive, and bone and dense contrast run into the hundreds or thousands.
- A monitor can only show a handful of grays at once, so we window: we pick a slice of the HU scale to display and throw the rest away.
- Two knobs run windowing — window level (the center) and window width (how much range you show). Same scan, ten different looks.
A CT scanner measures a vast range of tissue densities, but your eyeball gives up after about a few dozen shades of gray. Hounsfield Units and windowing are the bridge between what the scanner knows and what you can actually see. Get this one idea and half of "the CT looks different on lung windows" stops being mysterious.
What a Hounsfield Unit actually is
When an X-ray beam crosses your body, tissues eat part of it — the radiologists call this attenuation. Dense stuff (bone) eats a lot; airy stuff (lung) eats almost none. CT measures that attenuation for every little cube of tissue, called a voxel.
But "raw attenuation" is an awkward number that depends on the machine. So we standardize it onto a clean scale named after Sir Godfrey Hounsfield, who built the first clinical CT scanner. The trick is to anchor the scale to two things that are the same everywhere on Earth:
- Water = 0 HU
- Air = −1000 HU
Every other tissue gets a number relative to those two anchors. Think of it like sea level: we didn't invent where zero is, we just agreed water gets to be the reference line, and now everything is "above" or "below" it.
Water is 0 and air is −1000 by definition. These two points are baked into the HU scale itself, which is why a HU value means the same thing on a scanner in Tokyo as one in Toronto.
The neighborhood: who lives at what number
You don't need to memorize a table to the decimal, but a feel for the landmarks is genuinely useful — it's how you tell a fatty lump from a watery cyst from a bleed without guessing.
| Tissue | Rough HU | Why it's useful |
|---|---|---|
| Air (lung, bowel gas) | around −1000 | Pure black; the obvious stuff. |
| Fat | roughly −50 to −100 | Negative numbers basically mean fat — a great clincher for a fatty mass. |
| Simple fluid / water | near 0 | A cyst full of clean fluid sits close to zero. |
| Soft tissue / muscle | roughly +20 to +60 | The big crowded middle of the scale. |
| Acute blood (clotted) | often higher than plain fluid | Fresh blood reads denser than simple water. |
| Bone, dense contrast, metal | hundreds to thousands | The bright extreme. |
Notice the spread: from about −1000 all the way up past +1000. That's well over two thousand distinct values. Hold that thought, because here's the problem.
Because these numbers are standardized, you can literally measure a structure. Drop a region-of-interest circle on a kidney lesion, read the HU, and you've turned "looks kind of gray" into "+8 HU — that's just fluid, it's a benign cyst." Numbers don't lie the way eyeballs do.
Why we have to "window"
Your monitor and your retina can only distinguish on the order of dozens of gray shades at once — call it the difference between a black T-shirt and a slightly-less-black T-shirt. The CT scan has thousands of values. If you tried to cram the whole −1000-to-+2000 range into the few grays you can perceive, fat, fluid, muscle, and fresh blood would all smear into one indistinguishable porridge.
So we cheat. We don't show the whole scale at once. We pick a window — a slice of the HU range — and stretch just that slice across black-to-white. Everything below the window goes pure black; everything above goes pure white; the interesting middle gets all the contrast.
It's exactly like adjusting exposure on a photo of a bright window and a dark room. You can expose for the room (and blow out the window to white) or expose for the window (and lose the room to black) — but not both in one shot. Same data, different slice shown.
The two knobs: level and width
Windowing has only two controls, and once you name them they're easy:
- Window level (WL) — the center of the range you're displaying. Slide it up to look at brighter tissues, down to look at darker ones.
- Window width (WW) — how wide a range you spread across the grays. Narrow width = high contrast over a small HU span (great for separating tissues that are close in density). Wide width = low contrast but you see a huge range at once.
A narrow window is a magnifying glass for contrast: it makes tiny HU differences pop, which is exactly why subtle gray-on-gray findings like an early stroke are read on a narrow "stroke window." A wide window does the opposite — it sacrifices subtlety to fit extremes like bone and air in the same picture.
This is why a radiologist scrolls through the same chest CT on a soft-tissue window, then a lung window, then a bone window. They aren't different scans — they're the same numbers viewed through three different slices of the scale, because no single window can show a rib and a faint lung nodule and a liver lesion well at the same time. Picking the right window for the right tissue is genuinely a skill, and it ties directly into resolution, noise, and contrast.
A finding that's "invisible" can simply be in the wrong window, not absent. A faint lung nodule vanishes on soft-tissue windows; a subtle bony lesion disappears on lung windows. Before you call a study normal, make sure you actually looked through the window built for that tissue.
Putting it together
Hounsfield Units give every voxel an honest, standardized number for how dense it is — anchored forever to water at 0 and air at −1000. Windowing is the practical follow-up: since you can't perceive thousands of grays, you choose a slice of that scale and stretch it across the display, using level to aim and width to control contrast.
If you remember nothing else: HU is the measurement, windowing is how you choose to look at it. The data never changes — only your window onto it does. And when something refuses to show up, your first move isn't panic, it's the window knob. (For when the numbers themselves get distorted — like a streak between two hips reading falsely dense — see artifacts by modality.)