Computers in Imaging

1
Computers in Imaging

• Robert Metzger, Ph.D.

2
Decimal Form (Base 10)

• In general, a positional numbering system encodes
  the numbers as anbn an-1bn-1 . . . a2b2
  a1b1 a0b0 (0 lt ai lt b, i 0,1,2,...,n), where
  the integer b gt 1 is the radix (or base) of the
  numbering system
• The leftmost digit is called the most significant
  digit, the rightmost the least significant digit
• Whenever it is not clear which base is being used
  either a subscript will be used to denote it or
  the base will be written in parentheses
• Decimal form (radix 10) 4210 (4x101)(2x100)

3
Binary Form (Base 2)

• Powers of 2 1, 2, 4, 8, 16, 32, 64, 128, 256,
  512, 1024...
• Binary form (radix 2) 1010102 (1x25)(0x24)
  (1x23)(0x22) (1x21)(0x20) 3210 810 210
  4210
• Other radices used in computing Octal (b8)
  Hexadecimal (b16 A10, F15)

4
Digital Representation of Data
4

• Bits, Bytes and Words
• Smallest unit of storage capacity 1 bit (binary
  digit1 or 0)
• Bits grouped into bytes 8 bits byte
• Word 16, 32 or 64 bits, depending on the
  computer system addressing architecture
• Computer storage capacity is measured in
• kilobytes (kB) - 210 bytes 1024 bytes ? a
  thousand bytes
• megabytes (MB) - 220 bytes 1024 kilobytes ? a
  million bytes
• gigabytes (GB) - 230 bytes 1024 megabytes ? a
  billion bytes
• terabytes (TB) - 240 bytes 1024 gigabytes ? a
  trillion bytes

5
Digital Representation of Data
5

• Digital Representation of Different Types of Data
  
• Alphanumeric text, integers, and non-integer data
  
• Storage of Positive Integers
• In general, n bits have 2n possible permutations
  and can represent integers from 0 to 2n-1 (the
  range usually denoted with square brackets)
• n bits represents 2n values with range 0, 2n-1
• 8 bits represents 28 256 values with range 0,
  255
• 10 bits represents 210 1024 values with range
  0, 1023
• 12 bits represents 212 4096 values with range
  0, 4095
• 16 bits represents 216 65,536 values with range
  0, 65535

6
Digital Representation of Data
6

• Binary Representation of Signed Integers
• Include the use of negative numbers
• Reserve first bit for the sign (/-) -127,127
  ones complement
• Twos complement -128, 127 simplifies
  electronic circuitry
• Floating Point Form
• For very large or very small numbers (e.g., 6.023
  x 1023)
• Similar to scientific notation 0.111111112 x
  2010011112
• Binary Representation of Alphanumeric text
• ASCII American Standard Code for Information
  Interchange
• ASCII code for representation of text, e.g., A
  01000001
• Stored in one byte (128 characters)
• Computer needs to keep track of the data type

7
Data Transfer
7

• Data are transferred between the various
  components of the computer and with devices
  external to the computer in binary format
• A voltage of fixed value (e.g., 5V) is used to
  represent 1
• Another voltage value (e.g., 0V) is used to
  represent 0
• clock frequency 1/t (usually given in MHz or
  GHz)

• Changes between the voltage states occur through
  synchronization signals from the computers clock
  1 clock cycle the minimum time increment (t) at
  which a 1 ? 0 or 0 ? 1 transition can occur

8
Serial vs. Parrellel
8

• Serial - pulses transmitted one after another
  over single wire
• Parallel - All pulses transmitted simultaneously
  over several wires
• If N wires are used, parallel transmission is
  predominantly N times faster than serial
  transmission
• Bus a bundle of wires used for parallel data
  transfers

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 66.
9
Digital Data Transfer
9

• Each device connected to the bus is identified by
  an address or a range of addresses
• Only one device at a time can transmit data on
  the bus
• In general only one device receives the
  transmitted data
• The sending device transmits receiving address
  data

• The width of a bus refers to the number of wires
  used to transmit data in parallel (e.g., 32 bits)
  A bus also contains wires for ground, control
  signaling, etc.

10
10
Analog and Digital Representation of Data

• Analog continuous waveform where the amplitude
  represents the numerical signal magnitude
• Advantages of digital
• resistance to accumulated errors
• error correction possible with the transmission
  of redundant information
• digital circuitry most often less expensive than
  analog
• Advantage of analog
• Often transmitted quicker

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 67.
11
11
Conversion of Analog Data to Digital Form

• The electronic measuring devices of medical
  scanners (e.g., transducers and detectors)
  produce analog signals
• Analog to digital conversion (analog to digital
  converter ADC)
• ADCs characterized by
• sampling rate or frequency (e.g., samples/sec 1
  MHz)
• number of bits output per sample (e.g., 12
  bits/sample 12-bit ADC)

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 69.
12
12
ADC Potential Loss of Data

• Sampling and quantization (digitization) loss of
  data (necessary evil) Minimum sampling frequency
  (Nyquist limit) to accurately represent signal
  Quantization error minimized through use of large
  number of bits/sample

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 69.
13
Back to the Future
13

• I think there is a world market for
• maybe five computers.
• - Thomas Watson, chairman of IBM, 1943

14
Components Function of a Digital Computer
14
Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., pp. 70 and 78.
15
Main Memory
15

• Random access memory (RAM) volatile
• Buffer between CPU and mass storage devices
• Memory addresses where data and instructions
  reside
• Also read-only memory (ROM) static
• DRAM dynamic RAM
• SRAM static RAM (cache)
• VRAM video RAM (display card)
• All RAM volatile!

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 71.
16
Central Processing Unit (CPU)
16

• CPU executes a sequence of instructions program
• A CPU contained on a single chip microprocessor
• A number of data storage locations storage
  registers
• Data
• Memory addresses
• Arithmetic Logic Unit (ALU)
• Logic operations and data transfer signaled via
  clock
• CPU speed measured in instructions or operations
  per second (e.g., MIPS or GFLOPS) and determined
  by
• CPU clock rate (e.g., MHz or GHz)
• Architecture (bits per instruction, e.g., 32-bit
  vs. 64-bit and parallel processing capabilities)

17
Central Processing Unit (CPU)
17

• CPU program execution
• A program is a sequence of instructions for CPU
  execution
• Instruction cycle - CPU fetches the instructions
  from memory and executes them sequentially
• An instruction may cause the CPU to perform one
  of the following
• Mathematical operation
• Transfer data
• Compare
• Jump to an instruction other than the next in the
  sequence
• Each instruction consists of two parts an opcode
  specifying the operation to be performed and an
  address

18
Input-Output (I/O) Bus and Expansion Slots
18

• Bus described under serial vs. parallel data
  transfer
• Most I/O buses are provided with expansion slots
  to accommodate printed circuit (PC) cards with
  multiple functions, e.g.
• Modem card ? modem and video display card ? video
  monitor
• Makes it possible to customize general-purpose
  computers for specific applications (e.g., MRI
  scanner) and to add additional functions and
  capabilities (e.g., ADC)
• I/O Ports serial, parallel, USB (Universal
  Serial Bus) and SCSI (Small Computer System
  Interface)

19
Mass Storage Devices
19

• Permit the non-volatile storage of programs and
  data
• Various formats based on
• Access time (e.g., msec or minutes) random or
  sequential
• Data transfer rate (e.g., kbps, Mbps or Gbps)
• Cost
• Portability
• Permanence (CD-R vs. CD-RW)
• All consist of
• Mechanical drive
• Storage medium
• Controller
• Hierarchical trade-off speed vs. cost per MB

20
Mass Storage Devices
20
Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 76.
21
Display Interface and Keyboard/Pointing Devices
21

• Display computer information in visual form
• Usually displayed on a video monitor or printed
• Cathode ray tube (CRT)
• Flat-panel display (TFT thin-film transistors)
• Video display controller/card
• Receive digital data from computer memory
• Store locally on card with VRAM (video RAM)
• Registers to manipulate the original image or
  text data
• DACs to convert into on-screen video image
• Usually computer equipped with keyboard, mouse,
  trackball or joystick (could be head-less though)

22
Acquisition and Communications Interface
22

• Acquisition interface ADC card(s), though more
  efficient for the modality electronics to perform
  ADC
• Computers also communications devices (PACS)
• Modem modulator/de-modulator (DAC - encoded
  signal on wire - ADC)
• Network interface card (NIC), e.g., Ethernet
• Needs unique address on the network
• Phone number of modem pool, e.g., 206-685-5599
• Internet Protocol (IP) address, e.g., 128.95.120.1

23
Array Processor
23

• In the past when general-purpose CPU speeds were
  slow, custom-designed hardware (array processors)
  to perform compute-intense mathematical
  operations (e.g., floating point computation)
  were manufactured Achieved speed through
  specially designed circuits to make use of
  parallel processing and pipelining operation
• Attaches to the computer bus for fast I/O
  operation
• Not needed as much these days with very fast
  general-purpose microprocessors with parallel
  processing capabilities inherent in some
  operating systems

24
Performance of Computer Systems
24

• Review
• Clock speed of the CPU, e.g., 3.4GHz Pentium 4
• Width and clock speed of the I/O between
• Memory hierarchy, dimensions and elements
• Access and transfer times of mass storage devices
  
• MIPS, MFLOPS and benchmark testing
• CPU architecture, e.g.. number of
  bits/instruction and parallelism

25
Computer Languages
25

• Machine Language
• Binary instructions to be executed by CPU
  requiring detailed knowledge of the particular
  computer
• 0110101010101001001010101010001011110011110110101.
  ..)
• High-Level Languages
• Program writing without detailed knowledge of the
  machine
• This program is translated into machine language
  via a compiler
• Include FORTRAN, Basic, Pascal, C, Java
• Requires an compiler or interpreter program to
  translate to binary

26
Heirarchy of Software
26

• Applications Software - programs to perform
  specific functions desired by the user
• May be written in either high-level or machine
  language
• Generally an executable program run by the OS
• Hopefully user-friendly, flexible and intuitive
  to use
• Operating System (OS) - the program that, after
  being initially loaded into the computer by a
  boot program, manages all the other programs in a
  computer
• On instruction to run a program, the OS copies it
  from mass storage to memory, initiates execution
  of the first instruction by the CPU, transfers
  control to the program and regains control on
  completion of the task
• Handles complex I/O tasks and sharing of
  resources
• Examples Windows, Mac OS, Linux, UNIX

27
Computer Security
27

• Goals
• Deny unauthorized persons access to data
• Protect programs and data from accidental or
  deliberate loss
• Data Backup
• Practicing Safe Computing
• Malicious programs exist, such as viruses, worms,
  Trojans, time bombs, and password grabbers
• Types of viruses executable file, boot sector
  and macro infectors
• Deny unauthorized users access to your system
• Good password selection (8-14 characters, not in
  the dictionary of any known language, mix of
  upper/lower case and numbers, and should contain
  at least one non-alphanumeric character, e.g., !,
  _at_, , , etc.)
• Firewall software/hardware, e.g., Zone Alarm or
  Black Ice
• Grant each user only sufficient privileges
  required to accomplish required tasks

28
Back to the Future
28

• Computers in the future may weigh
• no more than 1.5 tons.
• - Popular Mechanics, forecasting the relentless
  march of science, 1949.

29
Digital Storage of Images
29

• Usually stored as a 2D array of data, I(x,y)
  I(1,1), I(2,1), I(n,m-1), I(n,m)
• Typical matrices CT 512x512x12bits/pixel DR
  2048x2560x10 bits/pixel
• Total number of bytes/image pixels/image
  bits/pixel (1 byte/8 bits)
• aligned along byte boundaries, e.g., 12
  bits/pixel ? 16 bits/pixel

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 71.
30
Effect of Resolution and Bits per Pixel
30
8, 3, 2, 1 bits/pixel
10242, 642, 322, 162 matrices
Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 82.
Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 84.
31
Image Processing
31

• Addition or subtraction, e.g., digital
  subtraction angiography (DSA)
• Spatial filtering
• Smoothing (removing quantum mottle noise)
• Edge enhancement, e.g., computed radiography (CR)
  
• Reconstruction from projections
• Back-projection, e.g., computed tomography (CT),
  single photon and positron emission tomography
  (SPECT and PET)
• Fast Fourier Transform, e.g., magnetic resonance
  imaging (MRI)
• Calculation of physiological performance indices,
  e.g., nuclear medicine
• Generation and manipulation of volumetric data
  sets
• Image co-registration (fusion), e.g., CT and PET

32
Back to the Future
32

• I have traveled the length and breadth of this
  country and talked with the best people, and I
  can assure you that data processing is a fad that
  won't last out the year.
• - The editor in charge of business books for
  Prentice Hall, 1957.

33
Computer-Aided Detection
33

• Also known as computer-aided diagnosis
• Computer program that uses specific image
  processing algorithms and decision threshold
  parameters to detect features in an image likely
  to be of clinical significance in images
• Assist as a secondary reader to call attention to
  objects that might have been overlooked
• For example in mammography
• Masses
• Microcalcification clusters
• Architectural distortions

34
Image Display
34

• Conversion of a digital image matrix in the
  display card memory (VRAM) into an analog video
  signal using a digital to analog converter (DAC)
  Matrix digital values are scanned in raster
  fashion as a function of time which through the
  DAC provides a time-varying analog signal
• The time-varying analog video signal is input to
  a video monitor

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., pp. 86 and 90.
35
Gray-scale and Color Cathode Ray Tube Monitors
35

• Gray-scale monitors provide better range of
  brightness and dynamic range than COTS color
  monitors
• CRT elements and function
• Intensity of light is proportional to the
  electric current in the beam, which is determined
  by the analog voltage signal applied from the
  video card
• A color CRT uses three independent electron guns
  with tightly clustered red, green and blue
  phosphor regions

36
Flat Panel Monitors
36

• Most flat-panel monitors use liquid crystal
  display (LCD) technology
• When voltage is applied to the liquid crystal
  material it rotates incident polarized light
• This rotated light then passes through another
  polarizer (90º to the first) so that the input
  voltage modulates the intensity of fluorescent
  tube backlight
• Active matrix LCDs are also called thin-film
  transistor (TFT) displays

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 89.
37
Contrast Enhancement
37

• Although there are 10-bit DACs, the human visual
  system (HVS) can only distinguish 26-28 shades of
  gray Thus for a 12-bit CT image, only 256 shades
  of gray are visualized at any one time of the
  4096 levels stored Also, radiographic contrast
  may vary between objects, so there is a need to
  interactively alter image contrast Altering the
  contrast so that it is more optimal involves the
  operation of a translation table (or look-up
  table LUT) sitting between VRAM and the DAC,
  allowing displayed image contrast enhancement

38
Video and Level Controls
38

• Modification of the translation table causes
  changes in the displayed image brightness and
  contrast and is usually done through window
  (contrast) and level (brightness) controls (e.g.,
  under mouse control)
• In the example (below), the window is kept
  constant as the level is increased, causing the
  image to become darker and darker
• The narrower the window, the greater the
  displayed image contrast

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 92.
39
False Color Displays and Hardcopy Devices
39

• The amplitude of the signals generated in the
  production of radiographic images do not have
  inherent color information
• When color is used to display some aspect of the
  received signal then the resulting images are
  called false-color or pseudo-color images
• Example Doppler US and NM
• Multiple LUTs and DACs
• Hardcopy Devices - permit the recording of
  digital images on photographic film or paper,
  e.g., laser imager

Bushberg, et al., The Essential Physics of
Medical Imaging, 2nd ed., p. 92.
40
PACS and Teleradiology
40

• Picture Archiving and Communications Systems
• Teleradiology
• Standards
• ACR Standards for Teleradiology
• Digital Imaging and Communications in Medicine
  (DICOM)
• Networks for Image and Data Transfer
• Acquisition of Digital Images
• Storage of Images
• Data Compression
• Display of images for Interpretation and
  Consultation

41
ACR Standards for Teleradiology
41

• Teleradiology (transmission of images for viewing
  at sites remote from where they are acquired) and
  reporting back ACR published the first ACR Std
  for Teleradiology in 1994 with subsequent
  revisions in 1996, 1998 and 2002
  http//www.acr.org/departments/stand_accred/standa
  rds/pdf/teleradiology.pdfThe ACR Standard for
  Teleradiology document outlines the
  qualifications of personnel involved, equipment
  guidelines, licensing, credentialing, and
  liability, communication, quality control for
  teleradiology, quality improvement and has a
  listing of up to date references

42
DICOM (Digital Imaging in Communications and
Medicine
42

• Most important functions
• unambiguous definition of terms used
• define models of image communication
• agreed upon by those who adopt the standard
• Has become the predominant standard for the
  communication of medical images
• Takes into account existing standards for
  networks
• By necessity, written in dry language with a
  minimum of explanatory information (thousands of
  pages)
• Web resource http//medical.nema.org/

43
Local Area Network (LAN)
43

• Topology star, ring and bus
• Protocol Internet and Ethernet use TCP/IP
  Transport Control Protocol/Internet Protocol
• Media wire, fiber-optic and air
• Ethernet
• Shared bandwidth
• Switched full duplex
• 10, 100, 1000 and 10,000 Mbps/sec
• ATM/SONET
• Asynchronous Transfer Mode
• Synchronous Optical Network
• 155, 622 and 2,500 Mbps

44
Wide Area Network (WAN)
44
Internet LANs mesh connected with WANs all
using TCP/IP
45
Acquisition of Digital Images
45

• Film digitization and frame grabbers (old)
• DICOM modalities
• Computed Tomography (CT)
• Magnetic Resonance Imaging (MRI)
• Nuclear Medicine (including SPECT and PET)
• Ultrasound (US)
• Computed Radiography (CR)
• Digital Radiography (DR)
• Digital Fluoroscopy (DF)
• Mammography (one digital, but mostly film)

46
Storage of Images
46

• Data Storage Technologies (redundancy and backup)
  
• Hierarchical Storage Management (HSM) systems
• RAID redundant array of inexpensive disks
• Magneto-optic disk (MOD/EOD)
• WORM optical disks
• Digital Linear Tape (DLT) and other tape formats,
  e.g., D2, D3 ...
• Data Compression
• Lossless (compression ratio lt 41)
• Lossy (compression ratio gt 41)

47
Display of Images for Interpretation
47

• Large format raster-scanned CRT and LCD (flat
  panel)
• CRT/LCD more or less equivalent to film
• Lesser spatial resolution, greater contrast
  resolution (LUT)
• Pixel Resolution 1024 x 1280 (C), 1200 x 1600
  (B), 1728 x 2304 (A)
• Luminance 240-280 cd/m2 vs. 1713 cd/m2 (light
  box)
• Dynamic Range bit depth (16-bit per pixel frame
  buffer)
• Veiling Glare stray ambient light reduces
  contrast
• ROC Studies OK for primary diagnosis
• Flat panel monitors available.
• 1536 x 2048 pixels and 700 cd/m2

48
Standard Image Matrices
49
Standard Image Matrices
50
IMAGE MODALITY
IMAGE SIZE (MB)
CT
0.5
MRI
0.125
CHEST
10.0
CR RAD
8.6 - 10.2
GI FLUORO
0.25
ANGIO --DSA FLUORO
0.5 - 2.0
NUCLEAR MED
0.033
ULTRASOUND
0.25 (0.75 COLOR)
51
pat / yr / unit
images/ pat
GB / year
of all GB data
modality
units
CT
6
80
7,250
13.5
1,740
MRI
3
80
2,900
87
0.7
CHEST
4
2
14,000
1,120
8.7
BONE
15
6
5,800
5,220
40.4
GI
8
45
2,500
1,256
9.7
52
pat / yr / unit
images/ pat
GB / year
of all GB data
modality
units
4
700
1,500
1,800
ANGIO
13.9
2
8
1,200
192
IVP
1.5
13
1
5,800
754
RAD P
5.8
30
8
1,750
105
C-ARM
0.8
25
40
2,500
625
US
4.8
NUC MED
6
20 - 50
2,250
18
0.1
53
OPTICAL DISK DATA STORAGE
54
MAGNETIC DATA STORAGE
55
WHAT ABOUT DIGITAL MAMMO?

• TYPICAL IMAGE SIZE
• 1500 x 1500 x 12 x 2 B 60 MB / IMAGE
• 4 image / pat x 100 pat / day x 290 days / year
• 7000 GB / YEAR ? ENTIRE SYSTEM CAPACITY
• HOW CAN FOLDER BE MADE WITH MULTI-MODALITY IMAGES
  FOR PATIENTS
• LONG-TERM STORAGE RETRIEVAL DIFFERENT FROM
  REGULAR CLINICAL IMAGES

56
TRANSMISSION OF IMAGE DATA
57
DATA TRANSMISSION TIMES
58
CR CASSETTE PLATE

• INSPECT FOR EXTERNAL DAMAGE
• INSPECT PLATE
• CLEAN PLATE
• BAR CODE ON CASSETTE
• SMOOTH OPEN / CLOSE
• NO ARTIFACTS ON PLATE
• PLATE EDGES ALIGNMENT

59
FILM CHARACTERISTIC CURVE
60
DYNAMIC RANGE OF AN IMAGING SYSTEM
FILM-SCREEN SYSTEMS
MAX DENSITY 1000 ?R
50
MIN DENSITY 20 ?R
DIGITAL SYSTEMS LIKE CR
MAX VALUE 100,000 ?R
10,000
MIN VALUE 10 ?R
61
SPATIAL RESOLUTION
PIXELS / 2 x FoV (mm)
2500 PIXELS / 2 x 430 mm
2.9 LINE PAIRS PER mm
FILM HAS ABOUT 6 - 8 LP / mm
62
(No Transcript)
63
IMAGE RETRIEVAL SPEEDS

• DEPENDS UPON SYSTEM TRAFFIC
• SHOULD BE EVALUATED DURING ACCEPTANCE TESTING
• TYPICAL SPEED FOR CR IMAGES FROM RAID STORAGE
  ABOUT 5 -15 SECONDS PER IMAGE
• TYPICAL SPEED FOR CR IMAGES FROM DLT LONG TERM
  STORAGE ABOUT 5 - 10 MINUTES --- IF TAPE ON-LINE

64
MAJOR SYSTEM PROBLEMS

• FAILURE OF RAID DRIVES
• CASSETTE DAMAGE FROM USE WITH MOBILE X-RAY
• MONITOR INTENSITY DEGRADATION WITH AGE --- 1 - 2
  YEAR REPLACEMENT
• FAILURE OF CR ERASURE LIGHT SOURCES
• CR READER MECHANICAL MISALIGNMENTS
• NETWORK INTERRUPTS
• LOST OR DUPLICATE IMAGE DUE TO TYPOs
• OPERATOR LACK OF FAMILIARITY ISSUES

65
INITIAL AEC TECHNIQUE CHART SET-UP

• E.I. 1000 x LOG10 EXPOSURE in mR 2000
  (Kodak)
• INPUT TO CASSETTE TYPICALLY SET AT 1.0 mR for AEC
  ? 150 - 200 RELATIVE FILM-SCREEN SPEED
• WE USE 0.8 - 0.85 mR
• ADDED 1.0 mm Al FILTERS TO X-RAY TUBES
• RAISE CLINICAL TECHNIQUE ABOUT 10 kVp
• RESULTS (1) CLINICAL NOISE CONTRAST ABOUT THE
  SAME, (2) PATIENT DOSES EQUALENT TO 250 - 300
  FILM-SCREEN SPEEDS

66
66
27 Step PACS Shuffle