Chapter 2 - Digital Image Processing

Front 1

Film based imaging has been the workhorse of radiology ever since the discovery of

Back 1

x-rays, 1895

Front 2

However, what has resolved film's shortcomings?

Back 2

digital image processing

Front 3

Steps in the production of film based

Back 3

1. Patient exposed to predetermined amount of radiation needed to provide diagnostic image quality

2. Latent image is formed on film that is subsequently processed by chemicals in a processor to render the image visible

3. Processed image is then ready for viewing by the radiologist- who makes the diagnosis

Front 4

Limitations of film based imaging

Back 4

may result in poor image quality

Front 5

Too dark IF initial radiation exposure is too

Back 5

high - film is overexposed, processed image appears too dark, the radiologist can't make diagnosis from image

Front 6

Too light IF initial radiation exposure is too

Back 6

low - too light, cannot be used by radiologist

Front 7

As a radiation detector, film screen cannot show differences in tissue contrast that are

Back 7

less than 10%

Front 8

Limitations: Optical range and contrast are

Back 8

fixed and limited

Front 9

Limitations: Requires manual handling for

Back 9

archiving and retrieval

Front 10

Increased radiation exposure

Back 10

repeated image

Front 11

Film is not ideal for performing 3 basic functions of radiation

Back 11

1. detection

2. image display

3. image archiving

Front 12

What is the highest of all imaging modalities?

Back 12

spatial resolution

*main reason radiography has played a significant role in imaging patients throughout the years

Front 13

Generic digital imaging system:
Major components include the following

Back 13

1. Data acquisition

2. Image processing

3. Image display/storage/archiving

4. Image communication

Front 14

Data acquisition refers to

Back 14

a systematic method of collecting data from the patient

Front 15

Data acquisition components

Back 15

1. xray tube

2. digital image detector

Front 16

Data acquisition is the measurement of the

Back 16

linear attenuation coefficient of the x-ray beam by digital image detectors

Front 17

The detectors produce an electronic signal that is converted by

Back 17

the analog to digital converter in preparation for processing by the computer

Front 18

The output signal from the detectors is an

Back 18

electrical signal - an analog signal that varies continuously in time

Front 19

Because digital computer is used,

Back 19

analog signal must be converted into a digital signal (discrete units) for processing by a digital computer

Front 20

This conversion is performed by a

Back 20

ADC - analog to digital converter

Front 21

For projection digital CT, the data are the

Back 21

electron density of tissues which is related to the linear attenuation coefficient

Front 22

It is attenuation data that are

Back 22

collected for these imaging modalities

Front 23

Image processing is

Back 23

performed by a computer using the binary number system

Front 24

Image processing uses digital information acquired by the

Back 24

ADC during the data acquisition phase and processes it into a format usable for diagnosis

Front 25

The ADC sends the digital data for

Back 25

digital image processing by a digital computer

Front 26

Image processing takes an

Back 26

input digital image and processes it to produce an output digital image by using a binary # system

Front 27

Binary system operates with base

Back 27

2, 0, 1

Front 28

These two digits are referred to as

Back 28

binary digits or bits - bits not continuous, they are discrete

Front 29

Computers operate with binary #s

Back 29

0 and 1

Front 30

These operations can be used to

Back 30

1. reduce noise in the output image

2. enhance the sharpness of the input image

3. change the contrast of the input image

Front 31

Image Display/Storage/Communication

Back 31

uses a DAC to convert the processed image into a viewable image on the computer monitor

Front 32

The output of computer processing, the output digital image, must be

Back 32

first converted into an analog signal before it can be displayed on a monitor for viewing by the observer

Front 33

Information is stored and archived on

Back 33

magnetic data (magnet tapes/disks) carriers and laser optical disks (for retrospective viewing and manipulation)

Front 34

Information can be sent electronically via

Back 34

computer networks to the PACS

Front 35

History of digital image processing dates back to

Back 35

early 1960s

^When NASA was developing its lunar and planetary exploration program

Front 36

Digital image processing is a multidisciplinary subject that includes

Back 36

1. physics

2. math

3. engineering

4. computer science

Front 37

Image Formation and Representation

Back 37

1. analog signal

2. digital signal

Front 38

Analog signal

Back 38

1. example- sine wave or a continuous function

2. made up of a comprehensive gray scale

Front 39

Digital signal

Back 39

1. discrete function

2. represented by numbers that can be processed by a computer

Front 40

Castlemans theory: images are

Back 40

all of subjects

Front 41

Within the set of images there are other subsets

Back 41

1. visible images

2. optical images

3. noninvisible physical images

4. mathematical images

Front 42

Visible images

Back 42

paintings, drawings, photographs

Front 43

Optical images

Back 43

holograms

Front 44

Noninvisible physical images

Back 44

temperature, pressure, elevation maps

Front 45

Mathematical images

Back 45

continuous and discrete functions

Front 46

Castleman noted that

Back 46

only the digital images can be processed by the computer

Front 47

Analog images are

Back 47

continuous images

ex. black and while photograph of chest x-ray because it represents a continuous distribution of light intensity as a function of position on the radiograph

Front 48

Digital images are

Back 48

numerical representations or images of objects

Front 49

Formation requires a

Back 49

digital computer

Front 50

Data must be in a

Back 50

digital format

Front 51

ADC is crucial in

Back 51

converting continuous (analog) signals to digital

Front 52

Analog processing

Back 52

both the input image and output image are analog

Front 53

DIgital processing

Back 53

both the input image and output image are discrete

Front 54

Process

Back 54

a series of actions or operations leading to a desired result

Front 55

Digital image processing

Back 55

subjecting numerical representations of objects to a series of operations in order to obtain a desired result

Front 56

In image processing, it is necessary to convert

Back 56

an input image into an output image

Front 57

In cases where an analog image must be converted into digital data for input to the computer, a

Back 57

digitization system is required

Front 58

CT is based on a

Back 58

reconstruction process whereby a digital image is changed into a visible physical image

Front 59

Image Domains - images can be represented in two domains on the basis of

Back 59

how they are acquired

Front 60

The two image domains?

Back 60

1. Spatial location domain

2. Spatial frequency domain

Front 61

Spatial location domain

Back 61

1. images viewed by humans

2. radiography and CT acquire info in spatial location domain

Front 62

Spatial frequency domain

Back 62

1. MRI acquires info

Front 63

Small structures within an object (patient) produce

Back 63

high frequencies that represent the detail in the image

Front 64

Large structures produce

Back 64

low frequencies and represent contrast info in the image

Front 65

Digital image processing can transform

Back 65

one image domain into another image domain

Front 66

Fourier transformation

Back 66

mathematical calculation performed by a computer - mathematically rigorous

Front 67

The fourier transformation converts

Back 67

image data from the spatial location domain to the spatial frequency domain, or vice versa

Front 68

The major reason for doing this is to

Back 68

facilitate image processing that can enhance or suppress certain features of the image

Front 69

The fourier transform converts a function in the

Back 69

time domain to a function in frequency domain

Front 70

Fundamental parameters of a digital image's structure

Back 70

1. matrix

2. pixels

3. voxels

4. bit depth

Front 71

A digital image is made up of a 2D array of # called a

Back 71

matrix

Front 72

The matrix consists of

Back 72

columns (M) and rows (N) that define small square regions called picture elements or pixels

Front 73

The size of the image can be described as

Back 73

MxNxk bits

Front 74

When M=N

Back 74

the image is square

Front 75

Matrix size is also sometimes referred to as

Back 75

FOV

Front 76

Generally, the diagnostic digital images are

Back 76

rectangular in shape

Front 77

The operator selects the matrix size by choosing the

Back 77

FOV

Front 78

As matrix size increases, images require

Back 78

more processing time, more storage space, and take longer to transmit to remote locations

Front 79

Pixels

Back 79

make up the matrix, generally square

Front 80

Each pixel contains

Back 80

a # (discrete value) that represents a brightness level or tissue characteristic

Front 81

In CT, these numbers are related to the

Back 81

1. atomic number

2. mass density of the imaged tissues

Front 82

Pixel size=

Back 82

FOV/matrix size

Front 83

The larger the matrix,

Back 83

the smaller the pixel size (for the same FOV) and the better the resolution (spatial resolution)

Front 84

Voxel

Back 84

contraction for volume element

Front 85

Voxel - pixels that are representing

Back 85

information contained in a volume of tissue

Front 86

Such volume is referred to as a

Back 86

voxel

Front 87

Voxel info is converted into

Back 87

numerical values contained in the pixels and these # are assigned brightness levels

Front 88

The higher # represent

Back 88

high signal intensity from the detectors - shaded white (bright)

Front 89

The lower # represent

Back 89

low signal intensity - shaded dark (black)

Front 90

Bit depth

Back 90

number of bits per pixel

Front 91

Bit depth is represented by

Back 91

"k bits" in the formula M x N x k bits

Front 92

k bits=

Back 92

2k - each pixel will have 2^k gray level (density)

Front 93

Matrix size has an effect on the

Back 93

detail or spatial resolution of the image

Front 94

What can affect the spatial resolution and density resolution of an image?

Back 94

1. matrix size

2. pixel size

3. bit depth

Front 95

Larger matrix

Back 95

smaller pixel size, improved spatial resolution

Front 96

FOV decreases

Back 96

smaller pixel size, improved spatial resolution

Front 97

Increase bit depth

Back 97

increase contrast resolution

Front 98

Image digitization - primary objective

Back 98

to convert an analog image into numerical data for processing by a computer

Front 99

Image digitization consists of 3 distinct steps

Back 99

1. scanning

2. sampling

3. quantization

Front 100

1st step

Back 100

scanning

Front 101

Scanning

Back 101

picture image is divided into small regions, pixels, placed within rows and columns, matrix

Front 102

The matrix allows

Back 102

identification of each pixel by providing an address for that pixel

Front 103

Increase the # of pixels in the image matrix and the

Back 103

image becomes more recognizable and facilitates better perception of image detail

Front 104

Each small region of the picture is a

Back 104

picture element (pixel)

Front 105

Scanning results in a

Back 105

grid characterized by rows and columns

Front 106

Size of the grid depends on the

Back 106

# of pixels on each side of the grid

Front 107

2nd step

Back 107

sampling

Front 108

Sampling

Back 108

brightness of each pixel is measured in the entire image

Front 109

A small spot of light is projected onto the transparency and the transmitted light is

Back 109

detected by a photomultiplier tube and outputs an electrical (analog) signal

Front 110

The output of the photomultiplier tube is an

Back 110

electrical (analog) signal

Front 111

final step

Back 111

Quantization

Front 112

Quantization

Back 112

electrical signal obtained from sampling is assigned an integer (0, or +/- #) proportional to the strength of that signal

Front 113

The result is each pixel being assigned a

Back 113

gray level ranging 0-255 placed on a rectangular grid

Front 114

Number 0 representing

Back 114

black

Front 115

Numer 255 representing

Back 115

white

Front 116

Number 1-254 representing a shade

Back 116

of gray

Front 117

The gray scale is based on the

Back 117

volume of gray levels

Front 118

The result of quantization

Back 118

1. a digital image

2. an array of # representing the analog image that was scanned, sampled, quantized

Front 119

Analog to digital conversion

Back 119

responsible in converting analog signals to digital information

Front 120

2 important characteristics of the ADC

Back 120

1. speed

2. accuracy

Front 121

Speed

Back 121

inversely proportional to accuracy - the greater the accuracy, the longer the digitization process

*time taken to digitize the analog signal

Front 122

Accurary

Back 122

the more samples taken, the more accurate the representation of the digital image

Front 123

Too few samples will result in

Back 123

aliasing artifacts

Front 124

Accuracy refers to the

Back 124

sampling of the signal

Front 125

Aliasing artifacts appear as

Back 125

Moire pattern on the image

Front 126

Why digitize images?

Back 126

1. image enhancement

2. image restoration

3. image analysis

4. image compression

5. image synthesis

Front 127

Image enhancement

Back 127

the purpose is to generate an image that is more pleasing to the observer

Front 128

Image restoration

Back 128

the purpose is to improve the quality of the images that have distortions/degradations

Front 129

Image analysis

Back 129

allows measurements and statistics to be performed

Front 130

Image compression

Back 130

the purpose is to reduce the size of the image to decrease transmission time and reduce storage space

Front 131

Image synthesis

Back 131

create images from other images or non-image data

Front 132

Image processing techniques are based on three types of operations

Back 132

1. point operations

2. local operations

3. global operations

Front 133

Point operations

Back 133

1. gray level mapping

2. histogram modification

Front 134

Local operations

Back 134

1. area processes/group processes

2. spatial frequency filtering

Front 135

Global operations

Back 135

fourier transform - entire input image is used to compute the value of the pixel into the output image; uses filtering in the frequency domain rather than space domain

Front 136

Alternate image processing technique

Back 136

Geometric operations

Front 137

Geometric operations

Back 137

changes the position (spatial position or orientation) of the pixel

`

Front 138

Geometric operations result in

Back 138

the scaling and sizing of the images and image rotation/translation

Front 139

Gray level mapping

Back 139

1. uses LUT - which plots the output/input gray levels against each other

Front 140

Gray level mapping changes the

Back 140

brightness of the image

Front 141

Gray level mapping results in the

Back 141

enhancement of the display image

Front 142

Gray level mapping results in a

Back 142

modification of the histogram of the pixel values

Front 143

Histogram

Back 143

a graph of the pixels plotted as a function of the gray level

Front 144

Histogram created by

Back 144

observing the image matrix and creating a table of the # of pixels with a specific intensity value

Front 145

Histogram - plotting a graph of the

Back 145

# of pixels versus the gray levels

Front 146

A histogram indicates the overall

Back 146

brightness and contrast of an image

Front 147

Histogram modification

Back 147

technique of modifying the histogram causing the brightness and contrast of the image to be modified

Front 148

Wide histogram results in

Back 148

high contrast

Front 149

Narrow histogram results in

Back 149

low contrast

Front 150

Low range values image appears

Back 150

dark

Front 151

Higher range values image appears

Back 151

bright

Front 152

Local operations

Back 152

image processing operation in which the output image pixel value is determined from a small area of pixels around the corresponding input pixel

Front 153

Example of local operations

Back 153

spatial frequency filtering

Front 154

Spatial frequency filtering

Back 154

1. high spatial frequency

2. low spatial frequencys

Front 155

High spatial frequency

Back 155

brightness of an image changes rapidly with distance in the horizontal/vertical direction

Front 156

An image with smaller pixels has

Back 156

higher frequency info than an image with larger pixels

Front 157

Low spatial frequency

Back 157

brightness changes slowly or at a constant rate

Front 158

Spatial location filtering: Convolution

Back 158

the value of the output pixel depends on a group of pixels in the input image that surround the input pixel of interest - pixel P5

Front 159

Convolution is a general purpose algorithm that is a technique of

Back 159

filtering in the space domain

Front 160

THe new value is a

Back 160

weighted average

Front 161

Convolution kernel

Back 161

each pixel in the kernel is a weighting factor or convolution coefficient

-size of kernel 3x3 matrix

Front 162

Spatial frequency filtering

Back 162

1. high pass filtering

2. low pass filtering

3. unsharp (blurred) masking

Front 163

High pass filtering is known as

Back 163

edge enhancement or sharpness

Front 164

High pass filtering is intended to

Back 164

sharpen an input image in the spatial domain that appears blurred

Front 165

Low pass filtering is used for the

Back 165

goal of image smoothing

Front 166

Smoothing is intended to

Back 166

reduce noise and the displayed brightness levels of pixels; however, image detail is compromised

Front 167

Unsharp (blurred) masking

Back 167

uses the blurred image produced from the low pass filtering process and subtracts it from the original image to produce a sharp image

Front 168

A CT image exam consists of

Back 168

two images per exam

Front 169

Image compression

Back 169

the use of software and hardware techniques to reduce information by removing unnecessary data

Front 170

Image compression allows

Back 170

remaining information to be encoded, stored or transmitted in an archive or storage media such as a tape or disc

Front 171

Upon decompression, the information is

Back 171

decoded and is filled with a representation of the data that was removed during compression

Front 172

Types of image compression

Back 172

1. Lossless

2. Lossy

Front 173

Lossless compression

Back 173

1. reversible

2. no info loss in compressed image data

3. does not involve the process of quantization

Front 174

Lossy compression

Back 174

1. irreversible

2. provides high compression ratios

3. currently not used by radiologists due to possibility of misdiagnosis

Front 175

Lossy compression involves 3 steps

Back 175

1. image transformation

2. quantization

3. encoding

Front 176

Image synthesis overview

Back 176

MRI, CT, 3D imaging in radiology, virtual reality imaging in radiology

Front 177

Virtual reality

Back 177

a branch of computer science that immerses the users in a computer generated environment and allows them to interact with 3D scenes- virtual endoscopy

Front 178

In CT - the slice of the patient is divided into small regions (voxels) because

Back 178

dimensions of depth (slice thickness) is added to the pixel

Front 179

Image processing hardware - basic image processing system consists of several interconnected components

Back 179

1. data acquisition device

2. digitizer

3. image memory

4. DAC

5. internal image processor

6. host computer

Front 180

Data acquisition device

Back 180

the video camera

-in CT this would be represented by the x-ray tube and detectors and detector electronics

Front 181

Digitizer

Back 181

analog signal converted into digital form by the digitizer, or ADC

Front 182

Image memory

Back 182

the digitized image is held in storage for further processing; size of the memory depends on the image

Front 183

DAC

Back 183

digital imaging held in the memory can be displayed on TV monitor; monitors the work with analog signals to convert digital data to analog signals with a DAC

Front 184

Internal imaging processor

Back 184

responsible for high speed processing of the input digital data

Front 185

Host computer

Back 185

primary component capable of performing several functions; plays significant role in applications