Window level can show recorded detail, but it can't recover what was never recorded
Key takeaways
- Windowing can show any part of the recorded range.
- It can't add signal or remove noise from an underexposed image.
- Clipped or burned-out pixels are all forced to the maximum value and hold no information.
- It doesn't change patient dose, or an exposure indicator or deviation index worked out before display processing.
- Windowing can hide poor technique, so read the numbers.
Partly. Windowing can bring out detail the detector recorded but the default display didn’t show. It can’t recover detail that was never recorded.
What can windowing reveal?
Bushong gives the numbers. A 14-bit digital radiograph holds 16,384 gray values, and most people can separate about 30 shades. Window and level let you stretch any part of the recorded range across those shades. Small differences you couldn’t see can become visible, and Bushong ranks this contrast amplification as possibly digital imaging’s most important feature.
What can’t it recover?
| Problem | Can windowing fix it? | Why |
|---|---|---|
| Subtle contrast at the default display | Yes | The data is there |
| Underexposure | No | Too few photons means noise, and the noise stays |
| Clipped or burned-out anatomy | No | Those pixels are all at the maximum value |
| Anatomy outside the field | No | It isn’t in the image |
| Motion blur | No | The blur is recorded |
| Too much dose | No | The dose was already given |
AAPM Report 116 says digital exposure errors show up as noise changes. It also says clipped regions hold no information. Those clipped pixels all sit at one value, so windowing can’t separate them.
Why does this matter for technique?
Because windowing can hide a problem. AAPM Report 116 gives an example: a lumbar spine at too low a kVp can reach an acceptable deviation index, with the signs of poor penetration windowed away and the patient still taking extra dose. So don’t plan to fix technique later. Check the exposure indicator and the noise, and fix the next exposure. AAPM Report 116 also says over- and underexposure aren’t readily recognizable on digital images. By that reasoning, an AEC error on a small patient can be hidden the same way.
Does a bad image need a repeat?
Not always. AAPM Report 116 suggests repeating an overexposed image only if relevant anatomy is clipped or burned out. For underexposure it suggests consulting a radiologist below a DI of -1 and repeating below -3. These are suggested limits, and your department sets its own.
Where is this on the outline?
The ARRT® 2027 outline lists “brightness (e.g., window level)” and “contrast (e.g., window width)” under postprocessing, and “gross exposure error (e.g., loss of contrast, saturation)” under image evaluation. For the brightness direction, see window level direction.
More in the Image Production guide, the Digital Image Processing chapter and the Image Acquisition set.
Frequently asked questions
Can windowing fix an underexposed image?
No. It can make the image brighter or change contrast, but the noise stays because the photons never reached the detector. AAPM Report 116 suggests consulting a radiologist on a repeat below a DI of -1, and repeating below -3.
Can windowing fix burned-out anatomy?
No. When the converter saturates, all pixels in that region are forced to the maximum value and hold no information. AAPM Report 116 treats this as the reason to repeat an overexposed image.
So what can windowing reveal?
Detail the detector recorded but the default display didn't show. A 14-bit image holds 16,384 values and the eye separates about 30 shades, so windowing picks which range you see.
Does windowing change the deviation index?
Not on an indicator worked out before display processing, as AAPM Report 116 proposes. Some older vendor scales differ: the report says Canon's REX depends on the brightness and contrast settings. Either way, an overexposed image that looks perfect after windowing is still overexposed.
Sources
Found an error? Email [email protected] with the page and the source.
Know the answer. Now drill it.
Free tier includes 3 chapters, 50 practice questions and 1 full timed mock. No credit card.