Computer System
Basics
byte = 8 bits (0 or 1); kilo = 210 = 1024
integrated circuits = complex electronic circuitry compressed into small plastic packages
silicon chips = silicon/semiconductors with high V = 1, low V = 0
CPU (central processing unit) = microprocessor
- CU = control unit = interpretation and executions of instructions
- ALU = arithmetic/logic unit = data manipulation
- clock rate = measure of speed in computations per sec
- number of bits (32/64) determines how much data can be processed in parallel
RAM (random access memory) = short-term volatile memory
ROM (read only memory) = non-volatile permanent storage for basic aspects of computer operation
magnetic storage (tape, floppy disk, hard disk) stores by controlling local magnetism at different points
optical disks (CD [compact disk], DVD [digital versatile disk], blu-ray disk)
computer bus = communication pathway for components of computer
software
- operating system
- application packages
- programming packages: editor (writing text), library of subroutines (small programs for common functions), linker (link written program to subroutine library) and compiler (translate written program to digital format)
Digital Image Processor
for image manipulation (eg enhancement) and image analyses (eg extraction of indices)
imaging system produces primary image information, fed to image acquisition circuitry for ADC and manipulation of data with an input look-up table (eg for pre-processing)
image display circuitry uses DAC, manipulation of displayed images to enhance appearance using output look-up table, for image magnification, displaying multiple images, annotation of images
ALU for image data (eg image subtraction in DSA)
array processor performs manipulations at higher speeds (eg FFT, image reconstruction)
high speed bus also connects to image memory, storage and computer interface
Digital Imaging
analogue = time varying signal with pixels influencing each other
digital = pixel/voxel signals are separate in a matrix format
sampling = digitisation of spatial information by dividing into sampling points = picture elements = N x N array of pixels; N is a power of 2
quantisation = digitisation of brightness = pixel values, the range of which is gray scale = G shades of gray; G = 2m where m = number of bits required (256/8, 1024/10, 4096/12) where 0 = black and 1023 = white
number of bits required = b = N x N x m
lossless compression maintains quality whereas lossy reduces quality for better compression ratio
Image Processing
contrast enhancement = windowing = gray-level transformation of pixel values for display purposes using output look-up tables (LUT); level = threshold where lower pixel values displayed as black, window = threshold for white, window width = distance between white and black
kernels are matrices which are centred on each pixel and figures multiplied to the existing pixel values
sharpening – kernel (negative peripheral values) enhancing sharp discontinuities between pixels
edge enhancement – kernel to highlight edges
median filter = form of smoothing – replaces centre of sample with median
Gaussian filtering – replaces central pixel with weighted sum (in Gaussian distribution) of sample
erosion filter – replaces pixels with brightest of surrounding sample
dilation filter – replaces pixel with darkest of surrounding sample
shadow filter – places shadow of opposite density adjacent to structures
threshold filter – changes values to white or black depending on a threshold
Fourier Transform
FT converts image into a combination of sine waves with differing amplitudes (brightness), phases and frequencies (spatial frequency, generating positive and negatives values), ie frequency domain/space; can be plotted with amplitude vs spatial frequency (1D) or fy vs fx with different brightness (2D)
small objects with sharp edges require increasingly high spatial frequencies; a perfectly sharp edge has all spatial frequencies in equal proportion
frequency representation enables data manipulation (eg removal of noise at certain frequencies)
calculated numerically and quickly using the fast Fourier transform (FFT) rather than the slower analytical discrete Fourier transform (DFT)
inverse Fourier transform (iFT) transforms back to image domain/space
Picture Archival and Communication Systems (PACS)
system enabling access to data of imaging modalities and transfer at high speeds (gigabit LAN = local area network using TCP/IP) to remote viewing consoles or archival storage; connects to reporting/clinical workstations, web server, teleradiology server and laser imager
teleradiology sends images to another location to enable viewing elsewhere
images from modalities (incl film digitisers) are sent to modality servers with RAID (redundant array of independent disks) and online archive simultaneously
offline archive is the final long-term storage with magnetic tapes (slow but durable) with copies sent off-site for disaster recover
database servers keep track of image data, move and copy studies between servers, communicate with RIS/HIS and provide modality work-list services (passing orders to modalities)
requires standardisation of image data interchange (DICOM), interface with RIS (radiology information system) and thus HIS (hospital information system) to avoid duplications and erroneous patient data (using HL7 = hospital language 7 standard), high quality image displays, user friendly, efficient distribution of images with short transfer times
DICOM (digital imaging and communications in medicine) defines network transmission protocol, storage information (name, age, NHI etc) and monitor calibration and specification requirements
Diagnostic Monitors
colour LCDs have 3 subpixels/elements (red, green, blue)
medical grade LCDs have greater luminescence (Cd)
greyscale LCDs use these to modulate the greyscale to give more bits to the image; also have higher luminance due to lack of colour filter; require special separate video card
digital LCDs and images systems miss out the DAC step which would otherwise cause noise and less likely to pick up interference
human eye can resolve 1 minute of arc (1/60°) at 600mm ≈ pixel size of 0.175mm
JND (just noticeable difference) = minimum luminance change for average observer to notice a difference with a finite JNDs per luminance range (hence higher luminance = more JNDs); 10bit grey systems provide enough grey levels to access all the JNDs of the human eye
recommend minimum luminance of 170 Cd/m2 and min depth of 8 bits, 2MPX; ideally 500Cd/m2 and 10 bits, 3MPX
response curve = monitor’s unique luminance output to the input range of grey levels
monitors must be calibrated to the DICOM grey scale definition function (GSDF) due to human eye perception not linear, calibrated so perceived brightness corresponds to original greyscale data
Workstations
separate hard drives for operating system and data storage so applications do not slow the system
large memory to enable fast interactive viewing of images
Compression
can be compressed by 20% without any loss of data (lossless compared to lossy which reduces amount of information; takes time to uncompress)
wavelet compression (used by InteleViewer) breaks image into series of blocks (transformations) that overlay to give increasing resolution
Internet
modem (modulator/demodulator)
hub = connect anything to anything with many connections causing packet collisions
switches = connect network ports with traffic directed to recipients
routers = programmable and decide which way to send communications, connecting different networks to buffer and transfer data between
WWW (world wide web) = global assemblage of computer networks
TCP/IP (transmission control protocol/internet protocol) = standard internet protocol
SLIP (serial line internet protocol) and PPP (point-to-point protocol) for communication over phone line
FTP (file transfer protocol) for exchange of files
HTTP (hypertext transfer protocol) = for transfer of documents containing different media
DHCP (dynamic host configuration protocol) = automatic allocation of an IP address
POP (post office protocol), SMTP (simple mail transfer protocol) and NNTP (network news transfer protocol) for electronic mail and newsgroups
MIME (multipurpose internet mail extensions) = for incorporating audio and image files via e-mail
HTML (hypertext mark-up language) for writing HTTP documents
GIF (graphical interchange format), JPEG (joint photographical experts group) and PNG (portable network graphics) = image formats
MPEG (moving picture experts group ) – image sequences format
URL (uniform resource location) = address in format: transfer protocol://www.server.type of environment.country code/document name.document format
applet = small compiled application downloaded with HTML to run on client computer (eg Java)
CGI (common gateway interface) = scripts = software for server-side processing