Chapter 20 - digital image processing

Digital image terminology

A digital image is a ___ of ___

Matrix of pixels

Digital image terminology

Pixel-

And represents what

Contraction of “picture element”

Represents a “voxel” (volume element) of tissue

Digital image terminology

Field of view (FOV)

In imaging, can be defined as the collimated x-ray field

Digital image terminology

Spatial resolution-

Also called what

Image sharpness

ASL called detail, recorded detail, definition

Digital image terminology

Spatial resolution is also dependent on ___ size

Matrix

Digital image terminology

The larger the matrix, the ___ the spatial resolution

The higher the number of pixels, the ____ the spatial resolution

Higher and higher

Digital image terminology

Pixel pitch-

Distance from center of one pixel to the center of an adjacent pixel

Digital image terminology - pixel pitch

The smaller the pixel pitch, the ___ the spatial resolution

The larger the pixel size, the ___ the spatial resolution

Higher, lower

Which has more pixels, CT or X-ray

X-ray

X-ray= 3520 × 4280 pixels, CT= 512 × 512 pixels

How do you find pixel size

Physical dimension / matrix size

Finding pixel size example

What is the pixel size for a 250×250 mm image reconstructed at the display monitor on a 1024×1024 matrix?

250/1024 =0.244 mm pixel size → [244 micrometers)

Finding physical dimensions:

If the pixel size is 0.1 mm and the image matrix is 6000×6000, what are the physical dimension of the matrix?

6000 × 0.01 = 60 mm

So what would 60 mm x 60 mm

Each pixel is capable of representing a wide range of different shades of gray

• the number of shades of gray is determined by the grayscale ___

Bit depth

Bit depth-

The number of shades of gray possible for the hardware of the system to acquire & display

Bit depth formula -

Formula: # of grays = 2^n (n = bit depth)

Bit depth example

Bit depth of 8, how many shades of gray

2^8 =256 shades of gray

Bit depth

Dynamic range-

The range of IR exposure that the detector can use to create an image

Bit depth

What is the dynamic range formula

same as other

Dynamic range (# of shapes of gray) = 2^n ( n = bit depth)

Bit depth - dynamic range

____ have a very wide dynamic range, and also have a very wide exposure latitude

Digital image receptors

Bit depth

Gray scale-

The range of pixel values (or gray values) actually present on the displayed image

bit depth

Gray scale also sometimes called :

Contrast scale

Long gray scale-

Image has many shades of gray

Long gray scale - an image with a long gray scale has a ____ window width. ( and ___ contrast)

Wide window width and low contrast

Short gray scale-

Image has only a few different shades

Short gray scale: and image with a short gray scale has a ___ window width. ( and ___ contrast)

Narrow window width and high contrast

Digitizing an image

What are the 3 main steps

1. Scanning

2. Sampling

3. Quantization

Digitizing an image

1. Scanning-

Exposure field recognition ← segmentation

Digitizing an image: scanning

What is segmentation

Region of interest on IR is identified

• data (including pitch black from direct, raw exposure) must be eliminated

Digitizing an image

2. Sampling -

Analog exposure level of each pixel is measured

Digitizing an image

3. Quantization-

Exposure level of each pixel is assigned a number. Each number represents a gray value

Histogram construction-

Simply a matter of counting the number of pixels holding each gray level

Histogram construction

Histogram-

Image data is plotted as a graph

Signal intensity for each pixel is plotted

Histogram construction

Most common type of histogram has 2 lobes- what are they

Main lobe is tissue within the anatomy

The tail lobe is “raw” direct X-ray exposure

Histogram construction -

The point ___ must be identified so that any raw direct exposure is eliminated from any calculations

S-max

Histogram construction -

S-max:

Pixel value (S value) corresponding to the least radiopaque anatomy

• any S value above S-max would indicate raw/direct X-ray exposure

Histogram construction -

If a projection is chosen in which no direct raw IR exposure is likely, an algorithm for histogram analysis which does not search for ___ is used

s-max

Histogram construction

If the image is part of a barium study, an algorithm that searches for both _____ and _____ should be selected

S-min and s-max

Histogram construction

An ___ is calcuted and used to calculated the exposure indicator

S-average

Histogram construction

The pixel values between s-min and s-max are called the _____

Values of interest (VOI)

Automatic image rescaling / a:

Histogram adjustment

Automatic image rescaling:

Automatic rescaling-

Re-mapping the brightnesss and gray scale of the digital image to provide a consistent display of brightness and contrast, regardless of variations in exposure

• the step that automatically corrects brightness for each image

• A simple process of algebraically “re-labeling” incoming data

Automatic image rescaling:

The image histogram is compared to a ____ in the computer

Standardized histogram

Automatic image rescaling

Each body part has an ideal reference standardized histogram which is found in-

Look up table (LUT)

Automatic image rescaling

If histograms do not match (which is very likely) the computer will___

Manipulate the new image data

Automatic image rescaling

Look up table (LUT)-

A permanent table of pixel values that represents the ideal histogram for that anatomic projection —> Q values

Automatic image rescaling

Rescaling means that regardless of the specific input values, the outputs values are always ___

The same

• the computer CAN align the average brightness of the image

Automatic image rescaling

Additionally, by aligning the high and low points of S and Q ranges, the computer CAN roughly rescale the ____, thus the ___ as well

Gray scale, contrast

Automatic image rescaling

What the computer cannot do is ___

Force the exact, specific shape of the histograms to match

Histogram equalization-

A processing technique that spread out the most frequent pixel intensity values or stretches out the intensity range of the image

Histogram equalization

• By doing this ^ histogram equalization allows the images areas with ___ contrast to gain ____ contsrat

Lowerr, higher

Limits of processing

Rescaling can correct for slight _____

Under and over exposure

Limits of processing

Under exposure: if under exposure is too great, ___ could results

Quantum mottle (a graininess from excessive noise (insufficient signal-to-noise ratio)

Limits of processing

Overexposure: if there’s a gross overexposure ___ of the IR could result

Saturation

Displaying the digital image

After the image has been rescaled, ____ and ____ are controlled by varying numerical values of each pixel

Brightness and contrast

Displaying the digital Image

Window level-

Determines midpoint of brightness displayed

Changes pixel value (addition and subtraction)

• controls overall image brightness

Displaying the digital Image

Window width:

Determines the number of shades of gray that will be displayed

Changes pixel value (multiplication and division)

• controls image contrast

Comparing equipment:

Detective quantum efficiency (DQE):

The measure of a digital systems ability to absorb remnant X-rays & convert them to useful signals

Comparing equipment:

High DQE =

More efficient system

• less patient dose required to get an optimal image

Exposure indicators;

Deviation index (DI):

New standard exposure indicator

Deviation index (DI)

• negative numbers =

• Positive numbers =

• underexposed

• Overexposed

Digital image quality factors

Brightness-

Controlled by computer

• window level (WL)

Digital image quality factors

Contrast:

Controlled by computer

• window width (WW)

• (kVp controls penetrability)

Digital image quality factors

Resolution (spatial resolution) sharpness

• Monitor matrix size / pixel size

• Pixel pitch

• Patient positioning

Digital image quality factors

Resolution (contrast resolution)

Grayscale bit depth

Digital image quality factors

Noise;

Measured as signal to noise ratio

• signal = remnant beam photons

• Noise= undesirable signals from the digital system

• Increases noise → lowers contrast

Digital image quality factors

Noise : quantum mottle

• noise that is visible on the image

• From insufficient number of photons (insufficient signal)

• Can be corrected by increasing mas

Digital image quality factors

Exposure latitude:

Defined as range of IR exposure that produce a quality image at an appropriate patient exposure

Digital image quality factors

Exposure latitude: in digital imaging, the computer can bring differing IR exposure into an appropriate visual range densite:

Slight underexpsire

Overexposure

If overexposure still results in an acceptable Imae, there is the potential for techniques to trend higher, defined as:

Dose creep

Reminder: it is th radiographers professional and ethical responsibility to minimize patient dose, known as

ALARA

Field uniformity corrections- known as

Digital equipment calibration

Field uniformity corrections does wha

Evens out brigtness across the area of the image field. Tested by low exposure with no object

Field uniformity corrections can occur due to what

• individual detector elements (dels or dealers) on IR being manufactured with slight variations in sensitivity

• Multiple amplifiers may not be perfectly aligned

• Different lengths of wires from individual dels offer more or less resistance

• Variable scintilator thickness

Noise reduction for del (Drexel) drop-out

What can this result from

In DR systems, dead detector elements (dels) can result from failure of TFTs or switching transistors

Noise reduction for del (dexel) drop-out

What’s the most common way to correct these

Using a software kernel in which the values of the 8 pixels surrounding a dead pixel are averaged, then this value is inserted into the dead pixel

Noise reduction for del (dexel) dropout

Kernel-

A sub-matrix (or mini matrix) that is passes over the original image matrix executing a mathematical function

Noise reduction for del (dexel) drop out

What’s the steps of a mathematical kernel function

• sum values for 8 surrounding pixels

• Average those values

• Insert average into centered pixel